MetaOmics: analysis pipeline and browser-based software suite for transcriptomic meta-analysis

SUMMARY

The rapid advances of omics technologies have generated abundant genomic data in public repositories and effective analytical approaches are critical to fully decipher biological knowledge inside these data. Meta-analysis combines multiple studies of a related hypothesis to improve statistical power, accuracy and reproducibility beyond individual study analysis. To date, many transcriptomic meta-analysis methods have been developed, yet few thoughtful guidelines exist. Here, we introduce a comprehensive analytical pipeline and browser-based software suite, called MetaOmics, to meta-analyze multiple transcriptomic studies for various biological purposes, including quality control, differential expression analysis, pathway enrichment analysis, differential co-expression network analysis, prediction, clustering and dimension reduction. The pipeline includes many public as well as >10 in-house transcriptomic meta-analytic methods with data-driven and biological-aim-driven strategies, hands-on protocols, an intuitive user interface and step-by-step instructions.

AVAILABILITY AND IMPLEMENTATION

MetaOmics is freely available at https://github.com/metaOmics/metaOmics.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Transcriptome Characterization of Matched Primary Breast and Brain Metastatic Tumors to Detect Novel Actionable Targets

BACKGROUND

Breast cancer brain metastases (BrMs) are defined by complex adaptations to both adjuvant treatment regimens and the brain microenvironment. Consequences of these alterations remain poorly understood, as does their potential for clinical targeting. We utilized genome-wide molecular profiling to identify therapeutic targets acquired in metastatic disease.

METHODS

Gene expression profiling of 21 patient-matched primary breast tumors and their associated brain metastases was performed by TrueSeq RNA-sequencing to determine clinically actionable BrM target genes. Identified targets were functionally validated using small molecule inhibitors in a cohort of resected BrM ex vivo explants (n = 4) and in a patient-derived xenograft (PDX) model of BrM. All statistical tests were two-sided.

RESULTS

Considerable shifts in breast cancer cell-specific gene expression profiles were observed (1314 genes upregulated in BrM; 1702 genes downregulated in BrM; DESeq; fold change > 1.5, Padj < .05). Subsequent bioinformatic analysis for readily druggable targets revealed recurrent gains in RET expression and human epidermal growth factor receptor 2 (HER2) signaling. Small molecule inhibition of RET and HER2 in ex vivo patient BrM models (n = 4) resulted in statistically significantly reduced proliferation (P < .001 in four of four models). Furthermore, RET and HER2 inhibition in a PDX model of BrM led to a statistically significant antitumor response vs control (n = 4, % tumor growth inhibition [mean difference; SD], anti-RET = 86.3% [1176; 258.3], P < .001; anti-HER2 = 91.2% [1114; 257.9], P < .01).

CONCLUSIONS

RNA-seq profiling of longitudinally collected specimens uncovered recurrent gene expression acquisitions in metastatic tumors, distinct from matched primary tumors. Critically, we identify aberrations in key oncogenic pathways and provide functional evidence for their suitability as therapeutic targets. Altogether, this study establishes recurrent, acquired vulnerabilities in BrM that warrant immediate clinical investigation and suggests paired specimen expression profiling as a compelling and underutilized strategy to identify targetable dependencies in advanced cancers.

Abstract P3-01-12: Breast cancer orbital and periorbital metastases can be bilateral, are associated with invasive lobular histology, and can co-occur with brain metastases

Breast cancer is responsible for most ophthalmic metastases (OM) to the orbit and periorbit. OM are associated with poor prognosis, as 95% of affected patients subsequently die of breast cancer. Previous studies have demonstrated an increased propensity for invasive lobular carcinoma (ILC) to spread to the eye compared to invasive ductal carcinoma (IDC). ILC is less common than IDC, and it can exhibit an unusual pattern of metastatic spread, including to sites such as the ovaries and GI tract. We herein report our single academic institution experience with OM from breast cancer with respect to anatomical presentation, histology (lobular vs. ductal), treatment and survival.

We used the natural language processing platform TIES (Text Information Extraction System) to identify OM caused by primary breast cancer in a database of over 2.3 million patients with electronically stored pathology and radiology reports at the University of Pittsburgh Medical Center. Search terms in TIES were translated into ontologies consisting of the NCI Metathesaurus’ synonyms and abbreviations. The search was thus not reliant on specific words. We then reviewed clinical notes, pathology reports, radiology reports, therapeutic regimens, and outcome data for the cases identified through TIES. Data from the patients identified through TIES was also compared to a large institutional database featuring 1,366 patients with metastatic breast cancer (MBC). Histological slides from 3 patients were analyzed.

We identified 28 patients diagnosed with primary breast cancer between 1995 and 2016 and subsequent OM. Median age at diagnosis was 54, with a range of 28 to 77. ER, PR, and HER2neu status from the 28 patients with OM did not differ from other patients with MBC in our institutional database. The relative proportion of patients with ILC was significantly higher in the patient cohort with OM (32.1%) than in the metastatic institutional database (11.3%, p=0.007). Median OS in the OM cohort was 78.4 months; distant metastasis free survival (DMFS) was 34.2 months. These survival times did not differ significantly from those patients in the large institutional metastatic database. DMFS tended to be longer (35.05 months) for patients with ILC compared to IDC (23.34 months), supporting a tendency for late relapse. Additionally, after a diagnosis of first metastasis, median survival of patients with ILC (21.4 months) was significantly shorter than that of patients with IDC (55.2 months) (p=0.03). OM were the second most frequent site of first metastasis in the OM cohort after bony metastases. Median time to first OM was 46.7 months. Of the 9 patients who developed bilateral OM, 4 had ILC, 1 had IDC, 2 had a mixed ILC/IDC, and 2 had an unknown histology. We observed a significant co-occurrence of OM and central nervous system (CNS) metastases (p=0.018). Of 14 patients that developed OM and CNS metastases, only 3 were diagnosed with ILC compared to 9 patients diagnosed with IDC. 57.1% of the patients with OM received radiation therapy to the eye, and 25 patients received at least one line of systemic therapy. The histological analysis revealed an interesting case in which the primary tumor was of a mixed ILC/IDC subtype, while only ILC was present in the OM.

To our knowledge, our report of 28 patients is the largest analysis of the histological subtype of breast cancer OM. Through our focus on anatomical presentation, histological subtype, treatment, and survival we provide a broad overview of this rare complication of breast cancer. Our data suggests that OM from breast cancer can often impact both eyes, can be associated with CNS metastases, and are more frequent in patients with ILC than IDC.

Citation Format: Martin Blohmer, Li Zhu, Jennifer M Atkinson, Sushil Beriwal, Joshua L. Rodriguez-Lopez, Margaret Rosenzweig, George C Tseng, Peter C Lucas, Adrian V Lee, Steffi Oesterreich, Rachel C Jankowitz. Breast cancer orbital and periorbital metastases can be bilateral, are associated with invasive lobular histology, and can co-occur with brain metastases [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P3-01-12.

Abstract P2-16-26: Mixed invasive ductal and lobular carcinoma (IDC/L) behaves similarly to invasive lobular carcinoma (ILC) with regard to neoadjuvant chemotherapy response and metastatic dissemination

Mixed invasive ductal and lobular carcinoma (Mixed IDC/L) is a rare subtype (3-5%) of invasive breast cancer with elusive pathophysiology. This entity is characterized by a mixed population of both ductal and lobular components within an individual tumor. Few studies have been published to date which compare Mixed IDC/L to invasive ductal carcinoma (IDC) and invasive lobular carcinoma (ILC) from a histopathologic perspective. Available literature on this topic is sparse and has been conflicting with regards to outcomes, and no studies to date describe the response of Mixed IDC/L to neoadjuvant chemotherapy. Patients with ILC have been shown to have lower response rates to neoadjuvant chemotherapy as compared to patients with IDC, which in turn leads to lower rates of successful breast conserving surgery and higher rates of re-excision due to positive margins. We aimed to compare Mixed IDC/L to pure ILC with regards to response to neoadjuvant chemotherapy and the need for repeat surgical intervention.

We identified 26 patients with Mixed IDC/L and 113 patients with ILC who received neoadjuvant chemotherapy at our institution between 1990 – 2017. At baseline, the groups had a similar median age, as well as ER H-score, PR H-score, and Ki-67 index. There was no statistical difference in rates of pathologic complete response (pCR) or percent tumor volume reduction post therapy between the two groups. Similarly, the percent of patients that required re-excision was not statistically different. Interestingly, the metastatic pattern was similar between the groups and included sites of dissemination such as the peritoneal cavity and omentum, which are not common sites of metastasis for IDC.

These findings suggest that Mixed IDC/L tumors behave similarly to ILC with regard to their response to neoadjuvant chemotherapy and patterns of metastatic spread. These findings support a prominent role for the lobular component in this mixed subtype in driving biology, including response to neoadjuvant therapy and metastatic dissemination. Ongoing efforts are directed towards incorporating data from the IDC cohort, as well as evaluation of changes in histology, ER/PR H-scores, and Ki-67 levels as a result of therapy. These data may imply that Mixed IDC/L tumors may behave clinically more like ILC than IDC, but larger studies are needed to study this rare breast cancer subtype.

Citation Format: Azadeh Nasrazadani, Jennifer M Atkinson, Yujia Li, Priscilla F McAuliffe, Rachel C Jankowitz, Leisha A Emens, George C Tseng, Adrian V Lee, Norman Wolmark, Steffi Oesterreich, Peter C Lucas. Mixed invasive ductal and lobular carcinoma (IDC/L) behaves similarly to invasive lobular carcinoma (ILC) with regard to neoadjuvant chemotherapy response and metastatic dissemination [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P2-16-26.

METABOLITES ASSOCIATED WITH HIGH VERSUS LOW WALKING ABILITY AMONG COMMUNITY-DWELLING OLDER MEN AND WOMEN

Low walking ability is highly prevalent with advanced age and associated with a higher risk of major adverse health outcomes. Metabolomics may help better characterize differences among older adults with vastly different walking abilities and provide insight into altered metabolic processes underlying age-related declines in physical functioning. Here, we sought to identify metabolites associated with high versus low walking ability using a nested case-control study of 120 community-dwelling adults ages 79-95 (40% men, 10% black) from the Cardiovascular Health Study (CHS) All Stars study. Participants with high versus low walking ability were matched one-to-one on age, gender, race, and fasting time. Using liquid chromatography-mass spectrometry, 569 metabolites were identified in overnight-fasting plasma. High versus low walking ability was defined as the best versus worst tertile of gait speed (≥0.9 versus <0.7 meters/second) and Walking Ability Index scores (7-9 versus 0-1). Ninety-six metabolites were associated with walking ability extremes (p<0.05, false discovery rate<30%), where 24% were triacylglycerols. Triacylglycerols containing mostly polyunsaturated fatty acids (e.g., omega-3) were higher, whereas those containing mostly saturated/monounsaturated fatty acids were lower among those with high versus low walking ability. Arginine and proline metabolism was a top pathway identified. Body mass index partly explained the association between a subset of metabolites and walking ability extremes. These findings may partly reflect pathways implicating modifiable risk factors including excess dietary lipids and lack of physical activity, which contribute to obesity and cause further alterations in metabolic pathways, potentially leading to age-related declines in walking ability in this cohort.

Metastatic breast cancers have reduced immune cell recruitment but harbor increased macrophages relative to their matched primary tumors

The interplay between the immune system and tumor progression is well recognized. However, current human breast cancer immunophenotyping studies are mostly focused on primary tumors with metastatic breast cancer lesions remaining largely understudied. To address this gap, we examined exome-capture RNA sequencing data from 50 primary breast tumors (PBTs) and their patient-matched metastatic tumors (METs) in brain, ovary, bone and gastrointestinal tract. We used gene expression signatures as surrogates for tumor infiltrating lymphocytes (TILs) and compared TIL patterns in PBTs and METs. Enrichment analysis and deconvolution methods both revealed that METs had a significantly lower abundance of total immune cells, including CD8+ T cells, regulatory T cells and dendritic cells. An exception was M2-like macrophages, which were significantly higher in METs across the organ sites examined. Multiplex immunohistochemistry results were consistent with data from the in-silico analysis and showed increased macrophages in METs. We confirmed the finding of a significant reduction in immune cells in brain METs (BRMs) by pathologic assessment of TILs in a set of 49 patient-matched pairs of PBT/BRMs. These findings indicate that METs have an overall lower infiltration of immune cells relative to their matched PBTs, possibly due to immune escape. RNAseq analysis suggests that the relative levels of M2-like macrophages are increased in METs, and their potential role in promoting breast cancer metastasis warrants further study.

Bayesian integrative model for multi-omics data with missingness

Motivation

Integrative analysis of multi-omics data from different high-throughput experimental platforms provides valuable insight into regulatory mechanisms associated with complex diseases, and gains statistical power to detect markers that are otherwise overlooked by single-platform omics analysis. In practice, a significant portion of samples may not be measured completely due to insufficient tissues or restricted budget (e.g. gene expression profile are measured but not methylation). Current multi-omics integrative methods require complete data. A common practice is to ignore samples with any missing platform and perform complete case analysis, which leads to substantial loss of statistical power.

Methods

In this article, inspired by the popular Integrative Bayesian Analysis of Genomics data (iBAG), we propose a full Bayesian model that allows incorporation of samples with missing omics data.

Results

Simulation results show improvement of the new full Bayesian approach in terms of outcome prediction accuracy and feature selection performance when sample size is limited and proportion of missingness is large. When sample size is large or the proportion of missingness is low, incorporating samples with missingness may introduce extra inference uncertainty and generate worse prediction and feature selection performance. To determine whether and how to incorporate samples with missingness, we propose a self-learning cross-validation (CV) decision scheme. Simulations and a real application on child asthma dataset demonstrate superior performance of the CV decision scheme when various types of missing mechanisms are evaluated.

Availability and implementation

Freely available on the GitHub at https://github.com/CHPGenetics/FBM.

Supplementary information

Supplementary data are available at Bioinformatics online.

Metabolites Associated with Vigor to Frailty Among Community-Dwelling Older Black Men

Black versus white older Americans are more likely to experience frailty, a condition associated with adverse health outcomes. To reduce racial disparities in health, a complete understanding of the pathophysiology of frailty is needed. Metabolomics may further our understanding by characterizing differences in the body during a vigorous versus frail state. We sought to identify metabolites and biological pathways associated with vigor to frailty among 287 black men ages 70-81 from the Health, Aging, and Body Composition study. Using liquid chromatography-mass spectrometry, 350 metabolites were measured in overnight-fasting plasma. The Scale of Aging Vigor in Epidemiology (SAVE) measured vigor to frailty based on weight change, strength, energy, gait speed, and physical activity. Thirty-seven metabolites correlated with SAVE scores (p < 0.05), while adjusting for age and site. Fourteen metabolites remained significant after multiple comparisons adjustment (false discovery rate < 0.30). Lower values of tryptophan, methionine, tyrosine, asparagine, C14:0 sphingomyelin, and 1-methylnicotinamide, and higher values of glucoronate, N-carbamoyl-beta-alanine, isocitrate, creatinine, C4-OH carnitine, cystathionine, hydroxyphenylacetate, and putrescine were associated with frailer SAVE scores. Pathway analyses identified nitrogen metabolism, aminoacyl-tRNA biosynthesis, and the citric acid cycle. Future studies need to confirm these SAVE-associated metabolites and pathways that may indicate novel mechanisms involved in the frailty syndrome.

Meta-analytic principal component analysis in integrative omics application

Motivation

With the prevalent usage of microarray and massively parallel sequencing, numerous high-throughput omics datasets have become available in the public domain. Integrating abundant information among omics datasets is critical to elucidate biological mechanisms. Due to the high-dimensional nature of the data, methods such as principal component analysis (PCA) have been widely applied, aiming at effective dimension reduction and exploratory visualization.

Results

In this article, we combine multiple omics datasets of identical or similar biological hypothesis and introduce two variations of meta-analytic framework of PCA, namely MetaPCA. Regularization is further incorporated to facilitate sparse feature selection in MetaPCA. We apply MetaPCA and sparse MetaPCA to simulations, three transcriptomic meta-analysis studies in yeast cell cycle, prostate cancer, mouse metabolism and a TCGA pan-cancer methylation study. The result shows improved accuracy, robustness and exploratory visualization of the proposed framework.

Availability and implementation

An R package MetaPCA is available online. (http://tsenglab.biostat.pitt.edu/software.htm).

Contact

ctseng@pitt.edu.

Supplementary information

Supplementary data are available at Bioinformatics online.

Meta-analytic framework for modeling genetic coexpression dynamics

Methods for exploring genetic interactions have been developed in an attempt to move beyond single gene analyses. Because biological molecules frequently participate in different processes under various cellular conditions, investigating the changes in gene coexpression patterns under various biological conditions could reveal important regulatory mechanisms. One of the methods for capturing gene coexpression dynamics, named liquid association (LA), quantifies the relationship where the coexpression between two genes is modulated by a third “coordinator” gene. This LA measure offers a natural framework for studying gene coexpression changes and has been applied increasingly to study regulatory networks among genes. With a wealth of publicly available gene expression data, there is a need to develop a meta-analytic framework for LA analysis. In this paper, we incorporated mixed effects when modeling correlation to account for between-studies heterogeneity. For statistical inference about LA, we developed a Markov chain Monte Carlo (MCMC) estimation procedure through a Bayesian hierarchical framework. We evaluated the proposed methods in a set of simulations and illustrated their use in two collections of experimental data sets. The first data set combined 10 pancreatic ductal adenocarcinoma gene expression studies to determine the role of possible coordinator gene USP9X in the Hippo pathway. The second experimental data set consisted of 907 gene expression microarray Escherichia coli experiments from multiple studies publicly available through the Many Microbe Microarray Database website (http://m3d.bu.edu/) and examined genes that coexpress with serA in the presence of coordinator gene Lrp.

Abstract P5-04-21: FGFR4 is a novel druggable target for recurrent ER-positive breast cancers

Background

Breast cancer recurrence is a major clinical problem for estrogen receptor positive (ER+) disease, even decades after initial surgery. These long-term recurrences are a challenge for invasive ductal carcinoma (IDC), and are particularly frequent for the histological subtype of invasive lobular carcinoma (ILC). To study the long-term endocrine resistance seen in ILC patients, our lab recently generated six long-term estrogen deprivation (LTED) models of ILC cells and performed RNA-Sequencing to identify differentially expressed genes that ostensibly allow these cells to grow in the absence of estrogen. We overlapped these results with a previously published microarray dataset of tamoxifen-resistant cells, and found that FGFR4 is the most consistently overexpressed gene in the setting of acquired resistance to endocrine therapy in ILC cells. From a recent publication of RNA-Seq from other LTED models, FGFR4 RNA overexpression is also seen in all five IDC cell lines.

Hypothesis

FGFR4 is an important mediator of acquired endocrine resistance in breast cancer.

Methods

To study the role of FGFR4 in vitro, we used multiple shRNAs and specific small molecule inhibition for growth assays. To study the role of FGFR4 in de novo resistance to endocrine therapy, we collected 129 well curated ER+ ILC tumor specimens and performed gene expression analysis on the pre-treatment samples using a custom NanoString panel. To study the role of FGFR4 in acquired resistance, we collected over 50 pairs of primary-metastatic ER+ tumors and performed exon capture based RNA-Sequencing.

Results

FGFR4 inhibition decreases parental and LTED cell growth in classic 2D conditions and in colony formation assays. The LTED cells, with higher FGFR4 expression, are more sensitive to its inhibition. For the parental cells, combination FGFR4 and ER-targeting drugs results in synergistic decreases in growth. In our database of primary ILC clinical samples, increased expression of FGFR4 is predictive of shorter time to distant recurrence. Among primary-recurrent tumor pairs, FGFR4 is an outlier expression gain in 20/50 (40%), spanning all recurrence sites studied (i.e. local recurrences, and metastases to the brain, bone, ovaries, and GI tract). Finally, in analyzing large cohorts of metastatic tumors, there is a significant enrichment of hotspot FGFR4 mutations in tumors originating in the breast, with &gt;2% of metastatic ILC tumors containing such a mutation.

Conclusion/Future studies

FGFR4 may play an important role in de novo resistance to endocrine therapy in ILC and acquired resistance in both ILC and IDC. Ongoing studies include overexpression of wild-type and FGFR4 hotspot mutations in ILC and IDC cell lines to determine growth and metastatic phenotypes.

Citation Format: Levine KM, Ding K, Priedigkeit N, Sikora MJ, Tasdemir N, Zhu L, Tseng GC, Jankowitz RC, Dabbs DJ, McAuliffe PF, Lee AV, Oesterreich S. FGFR4 is a novel druggable target for recurrent ER-positive breast cancers [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P5-04-21.

Abstract P2-01-09: ESR1 mutations drive breast cancer metastasis by context-dependent alterations in adhesive and migratory properties

Background: Estrogen receptor alpha (ERα/ESR1) is mutated in 30-40% of endocrine resistant ER+ breast cancer. These mutations, primarily located in the ligand binding domain, are associated with worse outcome in patients, and preclinical studies have shown that they cause ligand independent growth. An open question is whether these mutations contribute to actual metastatic process, or merely endocrine resistance.

Methods: Using Y537S and D538G genome-edited MCF7 and T47D cells, 3D growth was assessed in ultralow attachment plates. Cell-cell adhesion was determined using calcein-labelled adhesion assay and quantitative microfluidic fluorescence microscope (qMFM). Collagen-based adhesion and spheroid invasion assays were used to test adhesive and invasive properties. Wound scratching, spheroid collective migration and Boyden chamber transwell assays were applied to monitor cell migratory phenotypes. Mutated ER cistromes were profiled using ChIP-sequencing. ESR1 mutations in clinical samples were characterized using ddPCR.

Results: Visual inspection of cells grown in suspension culture revealed more compressed multicellular spheroids in ESR1 mutant cells, indicative of increased cell-cell interactions. This observation was confirmed in both static and microfluidic conditions. This effect was more pronounced in MCF7 than T47D cells, correlating with increased expression of desmosome and gap junction genes. Pharmacological blockade of gap junctions decreased cell-cell adhesion. Decreased attachment and increased invasion to collagen were discerned in all mutant cell types. Further functional analysis identified alterations in the TIMP3-MMP axis causing these phenotypes. The cell-cell adhesion phenotypes were restricted to MCF7-Y537S/D538G and T47D-Y537S, whereas T47D-D538G cells showed significantly increased migration. A GSEA screen identified Wnt signaling as uniquely induced in this context, and combination treatment using the Wnt inhibitor LGK974 and Fulvestrant led to synergistic inhibition of migration. ChIP-seq identified mutation-specific cistromes with an overall increased ligand-independent ER binding. However, it did not reveal binding sites in any candidate metastases genes, suggesting secondary epigenetic mechanisms. The motif analysis revealed the enrichment of FOXA1 motifs in mutated ER cistromes except T47D-D538G cells. However, knockdown of FOXA1 induced significantly higher inhibition of T47D-D538G migration than Fulvestrant treatment alone, indicating a FOXA1-dominated mechanism. Collectively, these data show that ESR1 mutant cells gain metastatic properties, in addition to endocrine resistance. To prove this using clinical samples, we measured ESR1 mutations in a well-defined cohort of endocrine resistant local or distant recurrence. Significant enrichment of ESR1 mutations in distant (9/55) vs local (0/27) recurrences confirms critical role of mutant ERα in metastases.

Conclusion: Further analysis of context dependent changes in cell-cell adhesion and migration of ESR1 mutant cells might guide the design and development of drugs targeting ERα-mutant tumors, such as inhibitors of gap junction, FOXA1, MMP, and Wnt signaling pathways.

Disclosure: The authors declare no conflict of interest.

Citation Format: Li Z, Bahreini A, Levine KM, Wang P, Tasdemir N, Montanez MA, Sundd P, Wallace CT, Watkins SC, Chu D, Park BH, Hou W, Mooring MS, Zhu L, Tseng GC, Carroll JS, Atkinson JM, Lee AV, Oesterreich S. ESR1 mutations drive breast cancer metastasis by context-dependent alterations in adhesive and migratory properties [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P2-01-09.

RNASeqDesign: A framework for RNA-Seq genome-wide power calculation and study design issues

Massively parallel sequencing (a.k.a. next-generation sequencing, NGS) technology has emerged as a powerful tool in characterizing genomic profiles. Among many NGS applications, RNA sequencing (RNA-Seq) has gradually become a standard tool for global transcriptomic monitoring. Although the cost of NGS experiments has dropped constantly, the high sequencing cost and bioinformatic complexity are still obstacles for many biomedical projects. Unlike earlier fluorescence-based technologies such as microarray, modelling of NGS data should consider discrete count data. In addition to sample size, sequencing depth also directly relates to the experimental cost. Consequently, given total budget and pre-specified unit experimental cost, the study design issue in RNA-Seq is conceptually a more complex multi-dimensional constrained optimization problem rather than one-dimensional sample size calculation in traditional hypothesis setting. In this paper, we propose a statistical framework, namely "RNASeqDesign", to utilize pilot data for power calculation and study design of RNA-Seq experiments. The approach is based on mixture model fitting of p-value distribution from pilot data and a parametric bootstrap procedure based on approximated Wald test statistics to infer genome-wide power for optimal sample size and sequencing depth. We further illustrate five practical study design tasks for practitioners. We perform simulations and three real applications to evaluate the performance and compare to existing methods.

Comprehensive Phenotypic Characterization of Human Invasive Lobular Carcinoma Cell Lines in 2D and 3D Cultures

Invasive lobular carcinoma (ILC) is the second most common subtype of breast cancer following invasive ductal carcinoma (IDC) and characterized by the loss of E-cadherin-mediated adherens junctions. Despite displaying unique histologic and clinical features, ILC still remains a chronically understudied disease, with limited knowledge gleaned from available laboratory research models. Here we report a comprehensive 2D and 3D phenotypic characterization of four estrogen receptor-positive human ILC cell lines: MDA-MB-134, SUM44, MDA-MB-330, and BCK4. Compared with the IDC cell lines MCF7, T47D, and MDA-MB-231, ultra-low attachment culture conditions revealed remarkable anchorage independence unique to ILC cells, a feature not evident in soft-agar gels. Three-dimensional Collagen I and Matrigel culture indicated a generally loose morphology for ILC cell lines, which exhibited differing preferences for adhesion to extracellular matrix proteins in 2D. Furthermore, ILC cells were limited in their ability to migrate and invade in wound-scratch and transwell assays, with the exception of haptotaxis to Collagen I. Transcriptional comparison of these cell lines confirmed the decreased cell proliferation and E-cadherin-mediated intercellular junctions in ILC while uncovering the induction of novel pathways related to cyclic nucleotide phosphodiesterase activity, ion channels, drug metabolism, and alternative cell adhesion molecules such as N-cadherin, some of which were differentially regulated in ILC versus IDC tumors. Altogether, these studies provide an invaluable resource for the breast cancer research community and facilitate further functional discoveries toward understanding ILC, identifying novel drug targets, and ultimately improving the outcome of patients with ILC.Significance: These findings provide the breast cancer research community with a comprehensive assessment of human invasive lobular carcinoma (ILC) cell line signaling and behavior in various culture conditions, aiding future endeavors to develop therapies and to ultimately improve survival in patients with ILC. Cancer Res; 78(21); 6209-22. ©2018 AACR.

Abstract B55: Recurrent transcriptional remodeling events and acquired fusion RNAs in relapsed ovarian cancers

Background: The majority of ovarian cancers (OvCa) are exquisitely sensitive to primary cytoreductive surgery and platinum-based agents, yet up to 80% of late-stage disease will relapse and develop deadly resistance to subsequent therapies. Identifying molecular mediators driving this acquired resistance is essential to improve the tragic 12- to 18-month prognosis for patients with recurrent disease. We hypothesize that relapsed ovarian cancers—which are largely uncharacterized—are molecularly distinct from primary disease and acquire druggable vulnerabilities throughout their life histories. To test this hypothesis, we undertook a transcriptome-wide analysis of 19 longitudinally collected patient-matched pairs of chemotherapy-naïve and recurrent cancers.

Materials and Methods: Ilumina TruSeq Total RNA-sequencing was performed on 19 flash-frozen patient-matched pairs of primary and recurrent OvCa. Time to recurrence was up to 64 months with a median of 25 months—a shared variant analysis confirmed all paired samples were patient-matched. Adapter-trimmed RNA-sequencing reads were quantified with k-mer based lightweight-alignment (Salmon v0.8.2) and transcript-abundance estimates were collapsed to gene-level with tximport. Differentially expressed genes were determined with DESeq2 using a paired model to account for patient-matched samples. To identify gains and losses in clinically actionable genes (DGIdb 2.0), pair-specific, outlier fold-change thresholds were defined as Q1/Q3 -/+ [1.5 X IQR], using each pairs’ expression fold-change values (recurrence vs. primary) as the distribution. These discrete, longitudinal transcriptional remodeling events (LTREs) in relapsed OvCa were then assessed for recurrence across all cases. Given that OvCa is thought to be driven largely by genomic structural variation, fusion RNAs were then called with FusionCatcher v0.99.7b. Identified fusions were filtered for cancer specificity by discarding fusions detected in normal tissue (Human Protein Atlas and BodyMap). The same fusion analysis was performed on CCLE OvCa cell line RNA-seq and selected fusions were validated with RT-PCR and Sanger sequencing.

Results: A suite of genes were consistently upregulated in OvCa recurrences, the most significant being NTRK2 (adjusted p-value &lt; 0.001)—a targetable tyrosine kinase. LTREs were also common with the most shared LTRE gains in recurrences being INHBA (44%) and IGF1 (39%). 18 of 19 (95%) recurrent cancers acquired cancer-specific fusion RNAs that were undetectable in the primary lesion. An in-frame, recurrence-acquired fusion between TOP2A, a target of doxorubicin and known chemoresistance mediator, and STAU1 was confirmed with RT-PCR. Lastly, we discovered recurrent (2 of 19 cases), in-frame CCDC6-ANK3 fusions that persisted throughout therapy in both the primary and relapsed lesions, each with distinct breakpoints. A CCDC6-ANK3 fusion was also validated in the chemoresistant OVCAR3 cell line. 15 of 19 cases (79%) harbored additional preserved fusions, albeit none were shared between cases.

Conclusions: Collectively, these results define multimodal transcriptomic mechanisms of ovarian cancer evolution in late disease. Considering that some acquisitions are highly recurrent and readily druggable (NTRK2, IHBA, IGF1), further preclinical studies are demanded and currently ongoing. Lastly, we establish acquired fusions involving known chemoresistance modulators and preserved fusion transcripts—which are maintained throughout therapy—as common somatic events in OvCa. Because fusion breakpoints are cancer specific, they may serve as promising patient-specific nucleotide targets and biomarkers.

Citation Format: Nolan Priedigkeit, Sarah Taylor, Shannon Grabosch, Jahnik Kurukulasuriya, Peter C. Lucas, Silvia Liu, Ester Elishaev, Amit Lugade, Kevin Eng, Anda Vlad, George C. Tseng, Kunle Odunsi, Robert P. Edwards, Adrian V. Lee. Recurrent transcriptional remodeling events and acquired fusion RNAs in relapsed ovarian cancers. [abstract]. In: Proceedings of the AACR Conference: Addressing Critical Questions in Ovarian Cancer Research and Treatment; Oct 1-4, 2017; Pittsburgh, PA. Philadelphia (PA): AACR; Clin Cancer Res 2018;24(15_Suppl):Abstract nr B55.

Improved identification of concordant and discordant gene expression signatures using an updated rank-rank hypergeometric overlap approach

Recent advances in large-scale gene expression profiling necessitate concurrent development of biostatistical approaches to reveal meaningful biological relationships. Most analyses rely on significance thresholds for identifying differentially expressed genes. We use an approach to compare gene expression datasets using ‘threshold-free’ comparisons. Significance cut-offs to identify genes shared between datasets may be too stringent and may miss concordant patterns of gene expression with potential biological relevance. A threshold-free approach gaining popularity in several research areas, including neuroscience, is Rank–Rank Hypergeometric Overlap (RRHO). Genes are ranked by their p-value and effect size direction, and ranked lists are compared to identify significantly overlapping genes across a continuous significance gradient rather than at a single arbitrary cut-off. We have updated the previous RRHO analysis by accurately detecting overlap of genes changed in the same and opposite directions between two datasets. Here, we use simulated and real data to show the drawbacks of the previous algorithm as well as the utility of our new algorithm. For example, we show the power of detecting discordant transcriptional patterns in the postmortem brain of subjects with psychiatric disorders. The new R package, RRHO2, offers a new, more intuitive visualization of concordant and discordant gene overlap.

Abstract PD4-09: Combination FGFR4 and ER-targeted therapy for invasive lobular carcinoma

Background

Invasive Lobular Carcinoma (ILC) is an understudied subtype of breast cancer that requires novel therapies in the advanced setting. Distinctive properties of ILC include growth patterns, metastatic behavior, and receptor status (almost universally estrogen receptor (ER) positive). Our lab recently generated six long-term estrogen deprivation (LTED) models of ILC cells and performed RNA-Sequencing to identify differentially expressed genes compared to their parental cells cultured with estrogen. We overlapped these results with a previously published microarray dataset and found that FGFR4 is the most consistently overexpressed gene in the setting of acquired resistance to endocrine therapy in ILC cells.

Hypothesis

FGFR4 is an important mediator of resistance to endocrine therapy in ILC.

Methods

To study the role of FGFR4 in vitro, we used multiple shRNAs and specific small molecule inhibition for growth assays of ILC cells. To study the role of FGFR4 in de novo resistance to endocrine therapy, we collected 129 well curated ER+ ILC tumor specimens and performed gene expression analysis on the pre-treatment samples using a custom NanoString panel. To study the role of FGFR4 in acquired resistance, we collected over 50 pairs of primary-metastatic ER+ tumors and performed exon capture based RNA-Sequencing.

Results

FGFR4 inhibition decreases parental and LTED ILC cell growth in classic 2D conditions, in the setting of ultra-low attachment, and in colony formation assays. The LTED cells, with higher FGFR4 expression, are more sensitive to its inhibition. For the parental cells, combination FGFR4 and ER-targeting drugs results in synergistic decreases in growth. In our database of primary ILC clinical samples, increased expression of FGFR4 is predictive of shorter time to distant recurrence. For our collection of 50 paired, primary-metastatic ER+ tissues, FGFR4 expression increases on average &gt;2.5 fold in the metastatic setting, with large gains even in ductal carcinoma cases. Finally, in analyzing recently published cohorts of metastatic tumors, there is a significant enrichment of hotspot FGFR4 mutations in tumors originating in the breast, with &gt;2% of metastatic ILC tumors containing such a mutation.

Conclusion

FGFR4 may play an important role in both acquired and de novo resistance to endocrine therapy in ILC.

Citation Format: Levine KM, Chen J, Sikora MJ, Tasdemir N, Priedigkeit N, Tseng GC, Puhalla SL, Jankowitz RC, Dabbs DJ, McAuliffe PF, Lee AV, Oesterreich S. Combination FGFR4 and ER-targeted therapy for invasive lobular carcinoma [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr PD4-09.

Abstract P1-03-03: Invasive lobular carcinoma and invasive ductal carcinoma differ in immune response, translation efficiency and metabolic rate

Background

Invasive lobular carcinoma (ILC) is the second most common histological subtype of breast cancer after invasive ductal carcinoma (IDC). ILC differs from IDC in pathologic, molecular, and clinical features. ILC tumors are most often characterized as luminal A by PAM50 analysis, suggestive of an indolent disease. Yet, when matched for receptor status and tumor grade, patients with ILC tend to have worse long-term outcomes than patients with IDC. The main distinguishing molecular feature of ILC is the loss of functional E-cadherin, and yet, beyond that loss, the mechanisms underlying the differences between ILC and IDC are largely unknown. We examined the RNA expression profiles of ILC and IDC tumors to assess if there may be underlying vulnerabilities of ILC tumors to novel therapeutic strategies.

Methods

Differential expression analysis was performed on 159 luminal A (LumA) ILC tumors versus 311 LumA IDC tumors from The Cancer Genome Atlas (TCGA). The METABRIC cohort (65 LumA ILC and 533 LumA IDC) was used as a validation dataset. Pathway enrichment analysis was performed to identify potential differences in biological processes, and these potential differences were then tested in a series of in vitro experiments, using 3 ER+ ILC (MDA-MB-134VI, SUM44PE, and MDA-MB-330) and 3 ER+ IDC (MCF7, T47D, and ZR75.1) cell lines.

Results

Pathway analysis led to the identification of three main signaling differences between LumA ILC and LumA IDC: immune regulation, translation, and metabolism. A series of immune pathways, including Immunological Synapse, Biocarta IL17 pathway, and Response to Wounding were up-regulated in ILC tumors. We examined specific cell type markers, and found that ILC tumors have a higher activity of nearly all immune cell types, including CD4+ T cells, CD8+ T cells, B cells, NK cells, dendritic cells, M1 macrophages, and M2 macrophages. These results were surprising, since ILC tumors have a lower incidence of stromal inflammation, as defined by H&E staining, suggesting a unique immune regulatory mechanism in ILC.

Next, we examined the translational regulation in ILC vs IDC tumors by comparing RNA expression and protein quantities as determined by RPPA analysis. ILC tumors have a lower protein:RNA ratio, suggesting a lower translation efficiency. This was reflected in the RPPA data by lower protein expression of eIF4G, ribosome protein S6 (S6) and p70-S6K in ILC tumors. Phosphorylation of 4E-BP1 (Ser65), eEF2, S6 (Ser235/236, Ser240/244), and mTOR (Ser2448) were also significantly lower in LumA ILCs. This lower translation efficiency was then validated in cell lines by O-propargyl-puromycin treatment.

Finally, the pathway analysis suggested lower rates of metabolism in lobular tumors. Comparative studies of OXPHOS and glycolysis with a Seahorse analyzer confirmed this finding.

Conclusions

ILC tumors have a higher immune activity than IDC tumors, even with lower rates of stromal inflammation, suggesting a potential for differential response to immunotherapy. The lower rates of translation and metabolism, which are general identifiers of tumor dormancy, could enable ILC to escape from cytotoxic therapies, and may play an important role in the late recurrence of ILC.

Citation Format: Levine KM, Du T, Zhu L, Tasdemir N, Lee AV, Van Houten B, Tseng GC, Oesterreich S. Invasive lobular carcinoma and invasive ductal carcinoma differ in immune response, translation efficiency and metabolic rate [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P1-03-03.

Sustained Molecular Pathology Across Episodes and Remission in Major Depressive Disorder

BACKGROUND

Major depressive disorder (MDD) is a debilitating mental illness and a major cause of lost productivity worldwide. MDD patients often suffer from lifelong recurring episodes of increasing severity, reduced therapeutic response, and shorter remission periods, suggesting the presence of a persistent and potentially progressive pathology.

METHODS

Subgenual anterior cingulate cortex postmortem samples from four MDD cohorts (single episode, n = 20; single episode in remission, n = 15; recurrent episode, n = 20; and recurrent episode in remission, n = 15), and one control cohort (n = 20) were analyzed by mass spectrometry-based proteomics (n = 3630 proteins) combined with statistical analyses. The data was investigated for trait and state progressive neuropathologies in MDD using both unbiased approaches and tests of a priori hypotheses.

RESULTS

The data provided weak evidence for proteomic differences as a function of state (depressed/remitted) or number of previous episodes. Instead it suggested the presence of persistent MDD effects, regardless of episodes or remitted state, namely on proteomic measures related to presynaptic neurotransmission, synaptic function, cytoskeletal rearrangements, energy metabolism, phospholipid biosynthesis/metabolism, and calcium ion homeostasis. Selected proteins (dihydropyrimidinase-related protein 1, synaptosomal-associated protein 29, glutamate decarboxylase 1, metabotropic glutamate receptor 1, and excitatory amino acid transporter 3) were validated by Western blot analysis. The findings were independent of technical, demographic (sex or age), or other clinical parameters (death by suicide and drug treatment).

CONCLUSIONS

Collectively, the results provide evidence for persistent MDD effects across current episodes or remission, in the absence of detectable progressive neuropathology.

Meta-analytic framework for liquid association

MOTIVATION

Although coexpression analysis via pair-wise expression correlation is popularly used to elucidate gene-gene interactions at the whole-genome scale, many complicated multi-gene regulations require more advanced detection methods. Liquid association (LA) is a powerful tool to detect the dynamic correlation of two gene variables depending on the expression level of a third variable (LA scouting gene). LA detection from single transcriptomic study, however, is often unstable and not generalizable due to cohort bias, biological variation and limited sample size. With the rapid development of microarray and NGS technology, LA analysis combining multiple gene expression studies can provide more accurate and stable results.

RESULTS

In this article, we proposed two meta-analytic approaches for LA analysis (MetaLA and MetaMLA) to combine multiple transcriptomic studies. To compensate demanding computing, we also proposed a two-step fast screening algorithm for more efficient genome-wide screening: bootstrap filtering and sign filtering. We applied the methods to five Saccharomyces cerevisiae datasets related to environmental changes. The fast screening algorithm reduced 98% of running time. When compared with single study analysis, MetaLA and MetaMLA provided stronger detection signal and more consistent and stable results. The top triplets are highly enriched in fundamental biological processes related to environmental changes. Our method can help biologists understand underlying regulatory mechanisms under different environmental exposure or disease states.

AVAILABILITY AND IMPLEMENTATION

A MetaLA R package, data and code for this article are available at http://tsenglab.biostat.pitt.edu/software.htm.

CONTACT

ctseng@pitt.edu.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Resilient protein co-expression network in male orbitofrontal cortex layer 2/3 during human aging

The orbitofrontal cortex (OFC) is vulnerable to normal and pathologic aging. Currently, layer resolution large-scale proteomic studies describing "normal" age-related alterations at OFC are not available. Here, we performed a large-scale exploratory high-throughput mass spectrometry-based protein analysis on OFC layer 2/3 from 15 "young" (15-43 years) and 18 "old" (62-88 years) human male subjects. We detected 4193 proteins and identified 127 differentially expressed (DE) proteins (p-value ≤0.05; effect size >20%), including 65 up- and 62 downregulated proteins (e.g., GFAP, CALB1). Using a previously described categorization of biological aging based on somatic tissues, that is, peripheral "hallmarks of aging," and considering overlap in protein function, we show the highest representation of altered cell-cell communication (54%), deregulated nutrient sensing (39%), and loss of proteostasis (35%) in the set of OFC layer 2/3 DE proteins. DE proteins also showed a significant association with several neurologic disorders; for example, Alzheimer's disease and schizophrenia. Notably, despite age-related changes in individual protein levels, protein co-expression modules were remarkably conserved across age groups, suggesting robust functional homeostasis. Collectively, these results provide biological insight into aging and associated homeostatic mechanisms that maintain normal brain function with advancing age.

Abstract 2841: Investigating drivers of disease progression in invasive lobular carcinoma

Invasive lobular carcinoma (ILC) is the second most common type of breast cancer following invasive ductal carcinoma (IDC) and accounts for 10-15% of all cases. Unlike the masses or lumps formed by IDCs, ILCs grow as small, dis-cohesive cells in a single-file pattern within dense layers of extracellular matrix. Paradoxically, while patients with ILC display favorable prognostic and predictive factors (ER+, PR+, HER2-, low Ki67), their long-term response to endocrine therapy is worse than patients with IDC. Despite such histological and clinical differences, ILC has remained a gravely understudied subtype of breast cancer. Thus, there is urgent need to investigate molecular drivers of ILC progression in order to develop more effective therapeutic strategies and improve patient outcome. To this end, we used the Nanostring platform to measure the expression of 577 copy number variation-associated genes in 131 primary ILC tumors and identified two candidate drivers that exhibit higher expression in tumors from patients with recurrent versus non-current disease: CTTN (cortical actin binding protein) and QSOX1 (quiescin sulfhydryl oxidase). In follow-up studies, we observed high CTTN and QSOX1 expression in human ILC cell lines, tumors and patient-derived xenografts. In addition, QSOX1 expression was further increased in ILC ovarian metastases compared to their matched primary tumors. In functional in vitro studies, RNAi-mediated inhibition of CTTN led to decreased adhesion and haptotaxis to Collagen I, while QSOX1 knockdown diminished the growth of human ILC cell lines. We are currently assessing the in vivo therapeutic utility of CTTN and QSOX1 inhibition in tumor growth and metastasis using human ILC cell line and patient-derived xenograft mouse models. This study greatly advances our understanding of ILC disease mechanisms and identifies novel therapeutic targets towards improving the clinical outcome of patients with this understudied subtype of breast cancer.

Citation Format: Nilgun Tasdemir, Matthew Sikora, Zhu Li, Kevin Levine, Ahmed Basudan, Soumya Luthra, Esther Elishaev, Uma Chandran, George C. Tseng, Rachel Jankowitz, David J. Dabbs, Priscilla McAuliffe, Nancy E. Davidson, Steffi Oesterreich. Investigating drivers of disease progression in invasive lobular carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 2841. doi:10.1158/1538-7445.AM2017-2841

A Joint Bayesian Model for Integrating Microarray and RNA Sequencing Transcriptomic Data

As the sequencing cost continued to drop in the past decade, RNA sequencing (RNA-seq) has replaced microarray to become the standard high-throughput experimental tool to analyze transcriptomic profile. As more and more datasets are generated and accumulated in the public domain, meta-analysis to combine multiple transcriptomic studies to increase statistical power has received increasing popularity. In this article, we propose a Bayesian hierarchical model to jointly integrate microarray and RNA-seq studies. Since systematic fold change differences across RNA-seq and microarray for detecting differentially expressed genes have been previously reported, we replicated this finding in several real datasets and showed that incorporation of a normalization procedure to account for the bias improves the detection accuracy and power. We compared our method with the popular two-stage Fisher's method using simulations and two real applications in a histological subtype (invasive lobular carcinoma) of breast cancer comparing PR+ versus PR- and early-stage versus late-stage patients. The result showed improved detection power and more significant and interpretable pathways enriched in the detected biomarkers from the proposed Bayesian model.

Age-Related Gene Expression in the Frontal Cortex Suggests Synaptic Function Changes in Specific Inhibitory Neuron Subtypes

Genome-wide expression profiling of the human brain has revealed genes that are differentially expressed across the lifespan. Characterizing these genes adds to our understanding of both normal functions and pathological conditions. Additionally, the specific cell-types that contribute to the motor, sensory and cognitive declines during aging are unclear. Here we test if age-related genes show higher expression in specific neural cell types. Our study leverages data from two sources of murine single-cell expression data and two sources of age-associations from large gene expression studies of postmortem human brain. We used nonparametric gene set analysis to test for age-related enrichment of genes associated with specific cell-types; we also restricted our analyses to specific gene ontology groups. Our analyses focused on a primary pair of single-cell expression data from the mouse visual cortex and age-related human post-mortem gene expression information from the orbitofrontal cortex. Additional pairings that used data from the hippocampus, prefrontal cortex, somatosensory cortex and blood were used to validate and test specificity of our findings. We found robust age-related up-regulation of genes that are highly expressed in oligodendrocytes and astrocytes, while genes highly expressed in layer 2/3 glutamatergic neurons were down-regulated across age. Genes not specific to any neural cell type were also down-regulated, possibly due to the bulk tissue source of the age-related genes. A gene ontology-driven dissection of the cell-type enriched genes highlighted the strong down-regulation of genes involved in synaptic transmission and cell-cell signaling in the Somatostatin (Sst) neuron subtype that expresses the cyclin dependent kinase 6 (Cdk6) and in the vasoactive intestinal peptide (Vip) neuron subtype expressing myosin binding protein C, slow type (Mybpc1). These findings provide new insights into cell specific susceptibility to normal aging, and suggest age-related synaptic changes in specific inhibitory neuron subtypes.

MetaDCN: meta-analysis framework for differential co-expression network detection with an application in breast cancer

MOTIVATION

Gene co-expression network analysis from transcriptomic studies can elucidate gene-gene interactions and regulatory mechanisms. Differential co-expression analysis helps further detect alterations of regulatory activities in case/control comparison. Co-expression networks estimated from single transcriptomic study is often unstable and not generalizable due to cohort bias and limited sample size. With the rapid accumulation of publicly available transcriptomic studies, co-expression analysis combining multiple transcriptomic studies can provide more accurate and robust results.

RESULTS

In this paper, we propose a meta-analytic framework for detecting differentially co-expressed networks (MetaDCN). Differentially co-expressed seed modules are first detected by optimizing an energy function via simulated annealing. Basic modules sharing common pathways are merged into pathway-centric supermodules and a Cytoscape plug-in (MetaDCNExplorer) is developed to visualize and explore the findings. We applied MetaDCN to two breast cancer applications: ER+/ER- comparison using five training and three testing studies, and ILC/IDC comparison with two training and two testing studies. We identified 20 and 4 supermodules for ER+/ER- and ILC/IDC comparisons, respectively. Ranking atop are 'immune response pathway' and 'complement cascades pathway' for ER comparison, and 'extracellular matrix pathway' for ILC/IDC comparison. Without the need for prior information, the results from MetaDCN confirm existing as well as discover novel disease mechanisms in a systems manner.

AVAILABILITY AND IMPLEMENTATION

R package 'MetaDCN' and Cytoscape App 'MetaDCNExplorer' are available at http://tsenglab.biostat.pitt.edu/software.htm .

CONTACT

ctseng@pitt.edu.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Combined systemic elimination of MET and epidermal growth factor receptor signaling completely abolishes liver regeneration and leads to liver decompensation

Receptor tyrosine kinases MET and epidermal growth factor receptor (EGFR) are critically involved in initiation of liver regeneration. Other cytokines and signaling molecules also participate in the early part of the process. Regeneration employs effective redundancy schemes to compensate for the missing signals. Elimination of any single extracellular signaling pathway only delays but does not abolish the process. Our present study, however, shows that combined systemic elimination of MET and EGFR signaling (MET knockout + EGFR-inhibited mice) abolishes liver regeneration, prevents restoration of liver mass, and leads to liver decompensation. MET knockout or simply EGFR-inhibited mice had distinct and signaling-specific alterations in Ser/Thr phosphorylation of mammalian target of rapamycin, AKT, extracellular signal-regulated kinases 1/2, phosphatase and tensin homolog, adenosine monophosphate-activated protein kinase α, etc. In the combined MET and EGFR signaling elimination of MET knockout + EGFR-inhibited mice, however, alterations dependent on either MET or EGFR combined to create shutdown of many programs vital to hepatocytes. These included decrease in expression of enzymes related to fatty acid metabolism, urea cycle, cell replication, and mitochondrial functions and increase in expression of glycolysis enzymes. There was, however, increased expression of genes of plasma proteins. Hepatocyte average volume decreased to 35% of control, with a proportional decrease in the dimensions of the hepatic lobules. Mice died at 15-18 days after hepatectomy with ascites, increased plasma ammonia, and very small livers.

CONCLUSION

MET and EGFR separately control many nonoverlapping signaling endpoints, allowing for compensation when only one of the signals is blocked, though the combined elimination of the signals is not tolerated; the results provide critical new information on interactive MET and EGFR signaling and the contribution of their combined absence to regeneration arrest and liver decompensation. (Hepatology 2016;64:1711-1724).

Integrative clustering of multi-level omics data for disease subtype discovery using sequential double regularization

With the rapid advances in technologies of microarray and massively parallel sequencing, data of multiple omics sources from a large patient cohort are now frequently seen in many consortium studies. Effective multi-level omics data integration has brought new statistical challenges. One important biological objective of such integrative analysis is to cluster patients in order to identify clinically relevant disease subtypes, which will form basis for tailored treatment and personalized medicine. Several methods have been proposed in the literature for this purpose, including the popular iCluster method used in many cancer applications. When clustering high-dimensional omics data, effective feature selection is critical for better clustering accuracy and biological interpretation. It is also common that a portion of "scattered samples" has patterns distinct from all major clusters and should not be assigned into any cluster as they may represent a rare disease subcategory or be in transition between disease subtypes. In this paper, we firstly propose to improve feature selection of the iCluster factor model by an overlapping sparse group lasso penalty on the omics features using prior knowledge of inter-omics regulatory flows. We then perform regularization over samples to allow clustering with scattered samples and generate tight clusters. The proposed group structured tight iCluster method will be evaluated by two real breast cancer examples and simulations to demonstrate its improved clustering accuracy, biological interpretation, and ability to generate coherent tight clusters.

Age-dependent increase in Kalirin-9 and Kalirin-12 transcripts in human orbitofrontal cortex

KALRN (KAL) is a Rho GEF that is highly involved in regulation of the actin cytoskeleton within dendrites. There are several isoforms of the protein that arise from differential splicing of KALRN's 66 exons. KAL isoforms have different functions in development. For example, overexpression of the KAL9 and KAL12 isoforms induce dendritic elongation in early development. However, in mature neurons KAL9 overexpression reduces dendritic length, a phenotype also observed in normal human ageing. We therefore hypothesized that KAL9 would have increased expression with age, and undertook to evaluate the expression of individual KALRN exons throughout the adult lifespan. Postmortem human brain grey matter from Brodmann's area (BA) 11 and BA47 derived from a cohort of 209 individuals without psychiatric or neurodegenerative disease, ranging in age from 16 to 91 years, were analysed for KALRN expression by Affymetrix exon array. Analysis of the exon array data in an isoform-specific manner, as well as confirmatory isoform-specific qPCR studies, indicated that the longer KAL9 and KAL12 isoforms demonstrated a statistically significant, but modest, increase with age. The small magnitude of the age effect suggests that inter-individual factors other than age likely contribute to a higher degree to KAL9 and KAL12 expression. In contrast to KAL9 and KAL12, global KALRN expression did not increase with age. Our work suggests that global measures of KALRN gene expression may be misleading and future studies should focus on isoform-specific quantification.

Statistical Methods in Integrative Genomics

Statistical methods in integrative genomics aim to answer important biology questions by jointly analyzing multiple types of genomic data (vertical integration) or aggregating the same type of data across multiple studies (horizontal integration). In this article, we introduce different types of genomic data and data resources, and then review statistical methods of integrative genomics, with emphasis on the motivation and rationale of these methods. We conclude with some summary points and future research directions.

Novel fusion transcripts associate with progressive prostate cancer

The mechanisms underlying the potential for aggressive behavior of prostate cancer (PCa) remain elusive. In this study, whole genome and/or transcriptome sequencing was performed on 19 specimens of PCa, matched adjacent benign prostate tissues, matched blood specimens, and organ donor prostates. A set of novel fusion transcripts was discovered in PCa. Eight of these fusion transcripts were validated through multiple approaches. The occurrence of these fusion transcripts was then analyzed in 289 prostate samples from three institutes, with clinical follow-up ranging from 1 to 15 years. The analyses indicated that most patients [69 (91%) of 76] positive for any of these fusion transcripts (TRMT11-GRIK2, SLC45A2-AMACR, MTOR-TP53BP1, LRRC59-FLJ60017, TMEM135-CCDC67, KDM4-AC011523.2, MAN2A1-FER, and CCNH-C5orf30) experienced PCa recurrence, metastases, and/or PCa-specific death after radical prostatectomy. These outcomes occurred in only 37% (58/157) of patients without carrying those fusion transcripts. Three fusion transcripts occurred exclusively in PCa samples from patients who experienced recurrence or PCaerelated death. The formation of these fusion transcripts may be the result of genome recombination. A combination of these fusion transcripts in PCa with Gleason's grading or with nomogram significantly improves the prediction rate of PCa recurrence. Our analyses suggest that formation of these fusion transcripts may underlie the aggressive behavior of PCa.

Polymorphisms in α-Defensin-Encoding DEFA1A3 Associate with Urinary Tract Infection Risk in Children with Vesicoureteral Reflux

The contribution of genetic variation to urinary tract infection (UTI) risk in children with vesicoureteral reflux is largely unknown. The innate immune system, which includes antimicrobial peptides, such as the α-defensins, encoded by DEFA1A3, is important in preventing UTIs but has not been investigated in the vesicoureteral reflux population. We used quantitative real-time PCR to determine DEFA1A3 DNA copy numbers in 298 individuals with confirmed UTIs and vesicoureteral reflux from the Randomized Intervention for Children with Vesicoureteral Reflux (RIVUR) Study and 295 controls, and we correlated copy numbers with outcomes. Outcomes studied included reflux grade, UTIs during the study on placebo or antibiotics, bowel and bladder dysfunction, and renal scarring. Overall, 29% of patients and 16% of controls had less than or equal to five copies of DEFA1A3 (odds ratio, 2.09; 95% confidence interval, 1.40 to 3.11; P<0.001). For each additional copy of DEFA1A3, the odds of recurrent UTI in patients receiving antibiotic prophylaxis decreased by 47% when adjusting for vesicoureteral reflux grade and bowel and bladder dysfunction. In patients receiving placebo, DEFA1A3 copy number did not associate with risk of recurrent UTI. Notably, we found that DEFA1A3 is expressed in renal epithelium and not restricted to myeloid-derived cells, such as neutrophils. In conclusion, low DEFA1A3 copy number associated with recurrent UTIs in subjects in the RIVUR Study randomized to prophylactic antibiotics, providing evidence that copy number polymorphisms in an antimicrobial peptide associate with UTI risk.

MetaKTSP: a meta-analytic top scoring pair method for robust cross-study validation of omics prediction analysis

MOTIVATION

Supervised machine learning is widely applied to transcriptomic data to predict disease diagnosis, prognosis or survival. Robust and interpretable classifiers with high accuracy are usually favored for their clinical and translational potential. The top scoring pair (TSP) algorithm is an example that applies a simple rank-based algorithm to identify rank-altered gene pairs for classifier construction. Although many classification methods perform well in cross-validation of single expression profile, the performance usually greatly reduces in cross-study validation (i.e. the prediction model is established in the training study and applied to an independent test study) for all machine learning methods, including TSP. The failure of cross-study validation has largely diminished the potential translational and clinical values of the models. The purpose of this article is to develop a meta-analytic top scoring pair (MetaKTSP) framework that combines multiple transcriptomic studies and generates a robust prediction model applicable to independent test studies.

RESULTS

We proposed two frameworks, by averaging TSP scores or by combining P-values from individual studies, to select the top gene pairs for model construction. We applied the proposed methods in simulated data sets and three large-scale real applications in breast cancer, idiopathic pulmonary fibrosis and pan-cancer methylation. The result showed superior performance of cross-study validation accuracy and biomarker selection for the new meta-analytic framework. In conclusion, combining multiple omics data sets in the public domain increases robustness and accuracy of the classification model that will ultimately improve disease understanding and clinical treatment decisions to benefit patients.

AVAILABILITY AND IMPLEMENTATION

An R package MetaKTSP is available online. (http://tsenglab.biostat.pitt.edu/software.htm).

CONTACT

ctseng@pitt.edu

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Abstract PD2-08: Potential role of prolactin signaling in development and growth of the lobular subtype of breast cancer

Invasive lobular carcinoma (ILC) is the eighth most frequently diagnosed cancer in any organ, and accounts for 8-11% of breast cancer. This histological subtype is characterized by loss of E-cadherin, and favorable prognostic factors, such as low Ki67 and high rates of ER/PR-positive tumors. Only recently is the lobular subtype gaining recognition as a distinct disease, displaying a unique growth pattern, unique molecular changes in addition to loss of E-cadherin, and evidence for late recurrences and reduced response to targeted endocrine therapy. It is widely accepted that a late age at first full term birth (FFTB) increases a women's risk for breast cancer. Interestingly, several published epidemiological studies have shown that the increased risk after a late age at FFTB is preferential for the lobular subtype of breast cancer compared to the ductal subtype. We therefore hypothesized that pregnancy hormones like prolactin play an integral role in the development and progression of ILC. Interrogation of the Cancer Genome Atlas (TCGA) data revealed a high expression of milk protein genes as well as prolactin signaling molecules, specifically Stat5a and Stat5b in lobular carcinomas compared to ductal carcinomas. We developed a lactation score including 7 milk protein genes and found that in the TCGA data set ILC tumors have a significantly higher lactation score than IDC tumors. Additionally, we found that ILC cell lines express increased prolactin receptor mRNA and protein levels compared to IDC cell lines. Prolactin treatment in ILC and IDC cells reveals divergent signaling pathways - prolactin stimulates ERK activation in IDC but not ILC cells. We are currently further delineating the prolactin signaling pathways, and resulting phenotypes, comparing ILC and IDC cells. We expect these experiments to move the field forward by establishing a relationship between prolactin and lobular carcinoma.

Citation Format: Oesterreich S, Katz TA, Logan G, Levine K, Nagle A, Huo Z, Tseng GC, Rui H, Lee AV, Butler LM. Potential role of prolactin signaling in development and growth of the lobular subtype of breast cancer. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr PD2-08.

A computational method for genotype calling in family-based sequencing data

Background

As sequencing technologies can help researchers detect common and rare variants across the human genome in many individuals, it is known that jointly calling genotypes across multiple individuals based on linkage disequilibrium (LD) can facilitate the analysis of low to modest coverage sequence data. However, genotype-calling methods for family-based sequence data, particularly for complex families beyond parent-offspring trios, are still lacking.

Results

In this study, first, we proposed an algorithm that considers both linkage disequilibrium (LD) patterns and familial transmission in nuclear and multi-generational families while retaining the computational efficiency. Second, we extended our method to incorporate external reference panels to analyze family-based sequence data with a small sample size. In simulation studies, we show that modeling multiple offspring can dramatically increase genotype calling accuracy and reduce phasing and Mendelian errors, especially at low to modest coverage. In addition, we show that using external panels can greatly facilitate genotype calling of sequencing data with a small number of individuals. We applied our method to a whole genome sequencing study of 1339 individuals at ~10X coverage from the Minnesota Center for Twin and Family Research.

Conclusions

The aggregated results show that our methods significantly outperform existing ones that ignore family constraints or LD information. We anticipate that our method will be useful for many ongoing family-based sequencing projects. We have implemented our methods efficiently in a C++ program FamLDCaller, which is available from http://www.pitt.edu/~wec47/famldcaller.html.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-016-0880-5) contains supplementary material, which is available to authorized users.

Regionally Smoothed Meta-Analysis Methods for GWAS Datasets

Genome-wide association studies are proven tools for finding disease genes, but it is often necessary to combine many cohorts into a meta-analysis to detect statistically significant genetic effects. Often the component studies are performed by different investigators on different populations, using different chips with minimal SNPs overlap. In some cases, raw data are not available for imputation so that only the genotyped single nucleotide polymorphisms (SNPs) results can be used in meta-analysis. Even when SNP sets are comparable, different cohorts may have peak association signals at different SNPs within the same gene due to population differences in linkage disequilibrium or environmental interactions. We hypothesize that the power to detect statistical signals in these situations will improve by using a method that simultaneously meta-analyzes and smooths the signal over nearby markers. In this study, we propose regionally smoothed meta-analysis methods and compare their performance on real and simulated data.

Comprehensive evaluation of fusion transcript detection algorithms and a meta-caller to combine top performing methods in paired-end RNA-seq data

Background: Fusion transcripts are formed by either fusion genes (DNA level) or trans-splicing events (RNA level). They have been recognized as a promising tool for diagnosing, subtyping and treating cancers. RNA-seq has become a precise and efficient standard for genome-wide screening of such aberration events. Many fusion transcript detection algorithms have been developed for paired-end RNA-seq data but their performance has not been comprehensively evaluated to guide practitioners. In this paper, we evaluated 15 popular algorithms by their precision and recall trade-off, accuracy of supporting reads and computational cost. We further combine top-performing methods for improved ensemble detection.

Results: Fifteen fusion transcript detection tools were compared using three synthetic data sets under different coverage, read length, insert size and background noise, and three real data sets with selected experimental validations. No single method dominantly performed the best but SOAPfuse generally performed well, followed by FusionCatcher and JAFFA. We further demonstrated the potential of a meta-caller algorithm by combining top performing methods to re-prioritize candidate fusion transcripts with high confidence that can be followed by experimental validation.

Conclusion: Our result provides insightful recommendations when applying individual tool or combining top performers to identify fusion transcript candidates.

Hypermethylation of BDNF and SST Genes in the Orbital Frontal Cortex of Older Individuals: A Putative Mechanism for Declining Gene Expression with Age

Expression of brain-derived neurotrophic factor (BDNF) and somatostatin (SST) mRNAs in the brain decreases progressively and robustly with age, and lower BDNF and SST expression in the brain has been observed in many brain disorders. BDNF is known to regulate SST expression; however, the mechanisms underlying decreased expression of both genes are not understood. DNA methylation (DNAm) is an attractive candidate mechanism. To investigate the contribution of DNAm to the age-related decline in BDNF and SST expression, the Illumina Infinium HumanMethylation450 Beadchip Array was used to quantify DNAm of BDNF (26 CpG loci) and SST (9 CpG loci) in the orbital frontal cortices of postmortem brains from 22 younger (age <42 years) and 22 older individuals (age >60 years) with known age-dependent BDNF and SST expression differences. Relative to the younger individuals, 10 of the 26 CpG loci in BDNF and 8 of the 9 CpG loci in SST were significantly hypermethylated in the older individuals. DNAm in BDNF exons/promoters I, II, and IV negatively correlated with BDNF expression (r=-0.37, p<0.05; r=-0.40, p<0.05; r=-0.24, p=0.07), and DNAm in SST 5' UTR and first exon/intron negatively correlated with SST expression (r=-0.48, p<0.01; r=-0.63, p<0.001), respectively. An expanded set of BDNF- and GABA-related genes exhibited similar age-related changes in DNAm and correlation with gene expression. These results suggest that DNAm may be a proximal mechanism for decreased expression of BDNF, SST, and other BDNF- and GABA-related genes with brain aging and, by extension, for brain disorders in which their expression is decreased.

Targeted DNA Methylation Screen in the Mouse Mammary Genome Reveals a Parity-Induced Hypermethylation of Igf1r That Persists Long after Parturition

The most effective natural prevention against breast cancer is an early first full-term pregnancy. Understanding how the protective effect is elicited will inform the development of new prevention strategies. To better understand the role of epigenetics in long-term protection, we investigated parity-induced DNA methylation in the mammary gland. FVB mice were bred or remained nulliparous and mammary glands harvested immediately after involution (early) or 6.5 months following involution (late), allowing identification of both transient and persistent changes. Targeted DNA methylation (109 Mb of Ensemble regulatory features) analysis was performed using the SureSelectXT Mouse Methyl-seq assay and massively parallel sequencing. Two hundred sixty-nine genes were hypermethylated and 128 hypomethylated persistently at both the early and late time points. Pathway analysis of the persistently differentially methylated genes revealed Igf1r to be central to one of the top identified signaling networks, and Igf1r itself was one of the most significantly hypermethylated genes. Hypermethylation of Igf1r in the parous mammary gland was associated with a reduction of Igf1r mRNA expression. These data suggest that the IGF pathway is regulated at multiple levels during pregnancy and that its modification might be critical in the protective role of pregnancy. This supports the approach of lowering IGF action for prevention of breast cancer, a concept that is currently being tested clinically.

The molecular landscape of premenopausal breast cancer

Introduction

Breast cancer in premenopausal women (preM) is frequently associated with worse prognosis compared to that in postmenopausal women (postM), and there is evidence that preM estrogen receptor-positive (ER+) tumors may respond poorly to endocrine therapy. There is, however, a paucity of studies characterizing molecular alterations in premenopausal tumors, a potential avenue for personalizing therapy for this group of women.

Methods

Using TCGA and METABRIC databases, we analyzed gene expression, copy number, methylation, somatic mutation, and reverse-phase protein array data in breast cancers from >2,500 preM and postM women.

Results

PreM tumors showed unique gene expression compared to postM tumors, however, this difference was limited to ER+ tumors. ER+ preM tumors showed unique DNA methylation, copy number and somatic mutations. Integrative pathway analysis revealed that preM tumors had elevated integrin/laminin and EGFR signaling, with enrichment for upstream TGFβ-regulation. Finally, preM tumors showed three different gene expression clusters with significantly different outcomes.

Conclusion

Together these data suggest that ER+ preM tumors have distinct molecular characteristics compared to ER+ postM tumors, particularly with respect to integrin/laminin and EGFR signaling, which may represent therapeutic targets in this subgroup of breast cancers.

Electronic supplementary material

The online version of this article (doi:10.1186/s13058-015-0618-8) contains supplementary material, which is available to authorized users.

Discovery and Classification of Fusion Transcripts in Prostate Cancer and Normal Prostate Tissue

Fusion transcript formation is one of the fundamental mechanisms that drives the development of prostate cancer. Because of the advance of high-throughput parallel sequencing, many fusion transcripts have been discovered. However, the discovery rate of fusion transcripts specific for prostate cancer is lagging behind the discoveries made on chromosome abnormalities of prostate cancer. Recent analyses suggest that many fusion transcripts are present in both benign and cancerous tissues. Some of these fusion transcripts likely represent important components of normal gene expression in cells. It is necessary to identify the criteria and features of fusion transcripts that are specific for cancer. In this review, we discuss optimization of RNA sequencing depth for fusion transcript discovery and the characteristics of fusion transcripts in normal prostate tissues and prostate cancer. We also propose a new classification of cancer-specific fusion transcripts on the basis of their tail gene fusion protein product and the roles that these fusions may play in cancer development.

Abstract P5-11-05: Pregnancy-induced epigenetic changes in the insulin-like growth factor signaling pathway

Prevention will prove to be the single most effective way of eradicating breast cancer. Currently, the most effective natural breast cancer prevention is an early first full term pregnancy. While it is not feasible to use pregnancy to protect women from breast cancer, understanding how the protective effect is elicited will inform the development of new prevention strategies. Women who were pregnant in their twenties are protected thirty to forty years later creating a complicated mechanism to tease out experimentally. In order to understand the long-term protection we have investigated epigenetics, specifically DNA methylation, which is known to be stable over long periods of time. A cohort (Parous) of female FVB mice were bred, gave birth, and pups were weaned. A control group (Nulliparpous) never saw male mice. Mammary glands were harvested immediately or 6 months after involution. These two time points allowed us to identify changes in DNA methylation that occurred in response to pregnancy, and additionally, changes that lasted long after parturition. DNA was isolated, and genome-wide DNA methylation was assessed using bisulfite-conversion and SureSelect Methyl-Seq. Bismark v0.7.12 was used for alignment of pair-end reads, followed by the R package "methylKit" for quality control and data analysis. A mapping efficiency of 50%∼68.1% was achieved with 89,512,619 base pairs covered. CpG Pearson correlation plots and PCA analysis showed global similarity between samples. We then conducted a logistic regression to ascertain parity-induced differentially methylated regions and identified 153 and 236 persistently hypomethylated and hypermethylated genes, respectively. Among the differentially methylated genes were many signaling molecules involved in growth factor signal transduction, including insulin-like growth factor 1 and 2 receptors (IGF1R and IGF2R). It has previously been shown that circulating IGF1 levels are reduced in parous women, and similarly the growth hormone/IGF axis is altered in rodent models. Collectively, these findings suggest that the IGF pathway is regulated at multiple levels during pregnancy, and that its modification might be critical in the protective role of pregnancy. We are currently following up on these data, including protein analysis of the IGF pathway members and downstream signaling molecules in human specimens. Finally, we are expanding our analysis to additional genes and pathways epigenetically altered by pregnancy, with the ultimate goal to develop new prevention strategies.

Citation Format: Tiffany A Katz, Serena G Liao, Thushangi Pathiraja, Robert K Dearth, George C Tseng, Steffi Oesterreich, Adrian V Lee. Pregnancy-induced epigenetic changes in the insulin-like growth factor signaling pathway [abstract]. In: Proceedings of the Thirty-Seventh Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2014 Dec 9-13; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2015;75(9 Suppl):Abstract nr P5-11-05.

Discovering monotonic stemness marker genes from time-series stem cell microarray data

Background

Identification of genes with ascending or descending monotonic expression patterns over time or stages of stem cells is an important issue in time-series microarray data analysis. We propose a method named Monotonic Feature Selector (MFSelector) based on a concept of total discriminating error (DE_total) to identify monotonic genes. MFSelector considers various time stages in stage order (i.e., Stage One vs. other stages, Stages One and Two vs. remaining stages and so on) and computes DE_total of each gene. MFSelector can successfully identify genes with monotonic characteristics.

Results

We have demonstrated the effectiveness of MFSelector on two synthetic data sets and two stem cell differentiation data sets: embryonic stem cell neurogenesis (ESCN) and embryonic stem cell vasculogenesis (ESCV) data sets. We have also performed extensive quantitative comparisons of the three monotonic gene selection approaches. Some of the monotonic marker genes such as OCT4, NANOG, BLBP, discovered from the ESCN dataset exhibit consistent behavior with that reported in other studies. The role of monotonic genes found by MFSelector in either stemness or differentiation is validated using information obtained from Gene Ontology analysis and other literature. We justify and demonstrate that descending genes are involved in the proliferation or self-renewal activity of stem cells, while ascending genes are involved in differentiation of stem cells into variant cell lineages.

Conclusions

We have developed a novel system, easy to use even with no pre-existing knowledge, to identify gene sets with monotonic expression patterns in multi-stage as well as in time-series genomics matrices. The case studies on ESCN and ESCV have helped to get a better understanding of stemness and differentiation. The novel monotonic marker genes discovered from a data set are found to exhibit consistent behavior in another independent data set, demonstrating the utility of the proposed method. The MFSelector R function and data sets can be downloaded from: http://microarray.ym.edu.tw/tools/MFSelector/.

Molecular and Genetic Characterization of Depression: Overlap with other Psychiatric Disorders and Aging

Genome-wide expression and genotyping technologies have uncovered the genetic bases of complex diseases at unprecedented rates; However despite its heavy burden and high prevalence, the molecular characterization of major depressive disorder (MDD) has lagged behind. Transcriptome studies report multiple brain disturbances but are limited by small sample sizes. Genome-wide association studies (GWAS) report weak results but suggest overlapping genetic risk with other neuropsychiatric disorders. We performed systematic molecular characterization of altered brain function in MDD, using meta-analysis of differential expression in eight gene array studies in three corticolimbic brain regions in 101 subjects. The identified "metaA-MDD" genes suggest altered neurotrophic support, brain plasticity and neuronal signaling in MDD. Notably, metaA-MDD genes display low connectivity and hubness in coexpression networks, and uniform genomic distribution, consistent with diffuse polygenic mechanisms. We next integrated these findings with results from over 1800 published GWAS and show that genetic variations nearby metaA-MDD genes predict greater risk for neuropsychiatric disorders and notably for age-related phenotypes, but not for other medical illnesses, including those frequently co-morbid with depression, or body characteristics. Collectively, the intersection of unbiased investigations of gene function (transcriptome) and structure (GWAS) provides novel leads to investigate molecular mechanisms of MDD and suggest common biological pathways between depression, other neuropsychiatric diseases, and brain aging.

VSNL1 Co-Expression Networks in Aging Include Calcium Signaling, Synaptic Plasticity, and Alzheimer's Disease Pathways

The visinin-like 1 (VSNL1) gene encodes visinin-like protein 1, a peripheral biomarker for Alzheimer disease (AD). Little is known, however, about normal VSNL1 expression in brain and the biologic networks in which it participates. Frontal cortex gray matter obtained from 209 subjects without neurodegenerative or psychiatric illness, ranging in age from 16 to 91, was processed on Affymetrix GeneChip 1.1 ST and Human SNP Array 6.0. VSNL1 expression was unaffected by age and sex, and not significantly associated with SNPs in cis or trans. VSNL1 was significantly co-expressed with genes in pathways for calcium signaling, AD, long-term potentiation, long-term depression, and trafficking of AMPA receptors. The association with AD was driven, in part, by correlation with amyloid precursor protein (APP) expression. These findings provide an unbiased link between VSNL1 and molecular mechanisms of AD, including pathways implicated in synaptic pathology in AD. Whether APP may drive increased VSNL1 expression, VSNL1 drives increased APP expression, or both are downstream of common pathogenic regulators will need to be evaluated in model systems.

Imputation of Truncated p-Values For Meta-Analysis Methods and Its Genomic Application

Microarray analysis to monitor expression activities in thousands of genes simultaneously has become routine in biomedical research during the past decade. a tremendous amount of expression profiles are generated and stored in the public domain and information integration by meta-analysis to detect differentially expressed (DE) genes has become popular to obtain increased statistical power and validated findings. Methods that aggregate transformed p-value evidence have been widely used in genomic settings, among which Fisher's and Stouffer's methods are the most popular ones. In practice, raw data and p-values of DE evidence are often not available in genomic studies that are to be combined. Instead, only the detected DE gene lists under a certain p-value threshold (e.g., DE genes with p-value < 0.001) are reported in journal publications. The truncated p-value information makes the aforementioned meta-analysis methods inapplicable and researchers are forced to apply a less efficient vote counting method or naïvely drop the studies with incomplete information. The purpose of this paper is to develop effective meta-analysis methods for such situations with partially censored p-values. We developed and compared three imputation methods-mean imputation, single random imputation and multiple imputation-for a general class of evidence aggregation methods of which Fisher's and Stouffer's methods are special examples. The null distribution of each method was analytically derived and subsequent inference and genomic analysis frameworks were established. Simulations were performed to investigate the type Ierror, power and the control of false discovery rate (FDR) for (correlated) gene expression data. The proposed methods were applied to several genomic applications in colorectal cancer, pain and liquid association analysis of major depressive disorder (MDD). The results showed that imputation methods outperformed existing naïve approaches. Mean imputation and multiple imputation methods performed the best and are recommended for future applications.

Missing value imputation in high-dimensional phenomic data: imputable or not, and how?

BACKGROUND

In modern biomedical research of complex diseases, a large number of demographic and clinical variables, herein called phenomic data, are often collected and missing values (MVs) are inevitable in the data collection process. Since many downstream statistical and bioinformatics methods require complete data matrix, imputation is a common and practical solution. In high-throughput experiments such as microarray experiments, continuous intensities are measured and many mature missing value imputation methods have been developed and widely applied. Numerous methods for missing data imputation of microarray data have been developed. Large phenomic data, however, contain continuous, nominal, binary and ordinal data types, which void application of most methods. Though several methods have been developed in the past few years, not a single complete guideline is proposed with respect to phenomic missing data imputation.

RESULTS

In this paper, we investigated existing imputation methods for phenomic data, proposed a self-training selection (STS) scheme to select the best imputation method and provide a practical guideline for general applications. We introduced a novel concept of "imputability measure" (IM) to identify missing values that are fundamentally inadequate to impute. In addition, we also developed four variations of K-nearest-neighbor (KNN) methods and compared with two existing methods, multivariate imputation by chained equations (MICE) and missForest. The four variations are imputation by variables (KNN-V), by subjects (KNN-S), their weighted hybrid (KNN-H) and an adaptively weighted hybrid (KNN-A). We performed simulations and applied different imputation methods and the STS scheme to three lung disease phenomic datasets to evaluate the methods. An R package "phenomeImpute" is made publicly available.

CONCLUSIONS

Simulations and applications to real datasets showed that MICE often did not perform well; KNN-A, KNN-H and random forest were among the top performers although no method universally performed the best. Imputation of missing values with low imputability measures increased imputation errors greatly and could potentially deteriorate downstream analyses. The STS scheme was accurate in selecting the optimal method by evaluating methods in a second layer of missingness simulation. All source files for the simulation and the real data analyses are available on the author's publication website.

Bias correction for selecting the minimal-error classifier from many machine learning models

MOTIVATION

Supervised machine learning is commonly applied in genomic research to construct a classifier from the training data that is generalizable to predict independent testing data. When test datasets are not available, cross-validation is commonly used to estimate the error rate. Many machine learning methods are available, and it is well known that no universally best method exists in general. It has been a common practice to apply many machine learning methods and report the method that produces the smallest cross-validation error rate. Theoretically, such a procedure produces a selection bias. Consequently, many clinical studies with moderate sample sizes (e.g. n = 30-60) risk reporting a falsely small cross-validation error rate that could not be validated later in independent cohorts.

RESULTS

In this article, we illustrated the probabilistic framework of the problem and explored the statistical and asymptotic properties. We proposed a new bias correction method based on learning curve fitting by inverse power law (IPL) and compared it with three existing methods: nested cross-validation, weighted mean correction and Tibshirani-Tibshirani procedure. All methods were compared in simulation datasets, five moderate size real datasets and two large breast cancer datasets. The result showed that IPL outperforms the other methods in bias correction with smaller variance, and it has an additional advantage to extrapolate error estimates for larger sample sizes, a practical feature to recommend whether more samples should be recruited to improve the classifier and accuracy. An R package 'MLbias' and all source files are publicly available.

AVAILABILITY AND IMPLEMENTATION

tsenglab.biostat.pitt.edu/software.htm.

CONTACT

ctseng@pitt.edu

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.