Detecting molecular subtypes from multi-omics datasets using SUMO

miss-SNF: a multimodal patient similarity network integration approach to handle completely missing data sources

MOTIVATION

Precision medicine leverages patient-specific multimodal data to improve prevention, diagnosis, prognosis, and treatment of diseases. Advancing precision medicine requires the non-trivial integration of complex, heterogeneous, and potentially high-dimensional data sources, such as multi-omics and clinical data. In the literature, several approaches have been proposed to manage missing data, but are usually limited to the recovery of subsets of features for a subset of patients. A largely overlooked problem is the integration of multiple sources of data when one or more of them are completely missing for a subset of patients, a relatively common condition in clinical practice.

RESULTS

We propose miss-Similarity Network Fusion (miss-SNF), a novel general-purpose data integration approach designed to manage completely missing data in the context of patient similarity networks. miss-SNF integrates incomplete unimodal patient similarity networks by leveraging a non-linear message-passing strategy borrowed from the SNF algorithm. miss-SNF is able to recover missing patient similarities and is "task agnostic", in the sense that can integrate partial data for both unsupervised and supervised prediction tasks. Experimental analyses on nine cancer datasets from The Cancer Genome Atlas (TCGA) demonstrate that miss-SNF achieves state-of-the-art results in recovering similarities and in identifying patients subgroups enriched in clinically relevant variables and having differential survival. Moreover, amputation experiments show that miss-SNF supervised prediction of cancer clinical outcomes and Alzheimer's disease diagnosis with completely missing data achieves results comparable to those obtained when all the data are available.

AVAILABILITY AND IMPLEMENTATION

miss-SNF code, implemented in R, is available at https://github.com/AnacletoLAB/missSNF.

MoNETA: MultiOmics Network Embedding for SubType Analysis

Cells are complex systems whose behavior emerges from a huge number of reactions taking place within and among different molecular districts. The availability of bulk and single-cell omics data fueled the creation of multi-omics systems biology models capturing the dynamics within and between omics layers. Powerful modeling strategies are needed to cope with the increased amount of data to be interrogated and the relative research questions. Here, we present MultiOmics Network Embedding for SubType Analysis (MoNETA) for fast and scalable identification of relevant multi-omics relationships between biological entities at the bulk and single-cells level. We apply MoNETA to show how glioma subtypes previously described naturally emerge with our approach. We also show how MoNETA can be used to identify cell types in five multi-omic single-cell datasets.

Integration of Multi-Omics Data for the Classification of Glioma Types and Identification of Novel Biomarkers

Glioma is currently one of the most prevalent types of primary brain cancer. Given its high level of heterogeneity along with the complex biological molecular markers, many efforts have been made to accurately classify the type of glioma in each patient, which, in turn, is critical to improve early diagnosis and increase survival. Nonetheless, as a result of the fast-growing technological advances in high-throughput sequencing and evolving molecular understanding of glioma biology, its classification has been recently subject to significant alterations. In this study, we integrate multiple glioma omics modalities (including mRNA, DNA methylation, and miRNA) from The Cancer Genome Atlas (TCGA), while using the revised glioma reclassified labels, with a supervised method based on sparse canonical correlation analysis (DIABLO) to discriminate between glioma types. We were able to find a set of highly correlated features distinguishing glioblastoma from lower-grade gliomas (LGGs) that were mainly associated with the disruption of receptor tyrosine kinases signaling pathways and extracellular matrix organization and remodeling. Concurrently, the discrimination of the LGG types was characterized primarily by features involved in ubiquitination and DNA transcription processes. Furthermore, we could identify several novel glioma biomarkers likely helpful in both diagnosis and prognosis of the patients, including the genes PPP1R8, GPBP1L1, KIAA1614, C14orf23, CCDC77, BVES, EXD3, CD300A, and HEPN1. Collectively, this comprehensive approach not only allowed a highly accurate discrimination of the different TCGA glioma patients but also presented a step forward in advancing our comprehension of the underlying molecular mechanisms driving glioma heterogeneity. Ultimately, our study also revealed novel candidate biomarkers that might constitute potential therapeutic targets, marking a significant stride toward personalized and more effective treatment strategies for patients with glioma.

Deep learning in cancer genomics and histopathology

Histopathology and genomic profiling are cornerstones of precision oncology and are routinely obtained for patients with cancer. Traditionally, histopathology slides are manually reviewed by highly trained pathologists. Genomic data, on the other hand, is evaluated by engineered computational pipelines. In both applications, the advent of modern artificial intelligence methods, specifically machine learning (ML) and deep learning (DL), have opened up a fundamentally new way of extracting actionable insights from raw data, which could augment and potentially replace some aspects of traditional evaluation workflows. In this review, we summarize current and emerging applications of DL in histopathology and genomics, including basic diagnostic as well as advanced prognostic tasks. Based on a growing body of evidence, we suggest that DL could be the groundwork for a new kind of workflow in oncology and cancer research. However, we also point out that DL models can have biases and other flaws that users in healthcare and research need to know about, and we propose ways to address them.

Proteomic Profile Distinguishes New Subpopulations of Breast Cancer Patients with Different Survival Outcomes

As a highly heterogeneous disease, breast cancer (BRCA) demonstrates a diverse molecular portrait. The well-established molecular classification (PAM50) relies on gene expression profiling. It insufficiently explains the observed clinical and histopathological diversity of BRCAs. This study aims to demographically and clinically characterize the six BRCA subpopulations (basal, HER2-enriched, and four luminal ones) revealed by their proteomic portraits. GMM-based high variate protein selection combined with PCA/UMAP was used for dimensionality reduction, while the k-means algorithm allowed patient clustering. The statistical analysis (log-rank and Gehan-Wilcoxon tests, hazard ratio HR as the effect size ES) showed significant differences across identified subpopulations in Disease-Specific Survival (p = 0.0160) and Progression-Free Interval (p = 0.0264). Luminal subpopulations vary in prognosis (Disease-Free Interval, p = 0.0277). The A2 subpopulation is of the poorest, comparable to the HER2-enriched subpopulation, prognoses (HR = 1.748, referenced to Luminal B, small ES), while A3 is of the best (HR = 0.250, large ES). Similar to PAM50 subtypes, no substantial dependency on demographic and clinical factors was detected across Luminal subpopulations, as measured by χ2 test and Cramér's V for ES, and ANOVA with appropriate post hocs combined with η2 or Cohen's d-type ES, respectively. Progesterone receptors can serve as the potential A2 biomarker within Luminal patients. Further investigation of molecular differences is required to examine the potential prognostic or clinical applications.

Identifying a distinct fibrosis subset of NAFLD via molecular profiling and the involvement of profibrotic macrophages

BACKGROUND

There are emerging studies suggesting that non-alcoholic fatty liver disease (NAFLD) is a heterogeneous disease with multiple etiologies and molecular phenotypes. Fibrosis is the key process in NAFLD progression. In this study, we aimed to explore molecular phenotypes of NAFLD with a particular focus on the fibrosis phenotype and also aimed to explore the changes of macrophage subsets in the fibrosis subset of NAFLD.

METHODS

To assess the transcriptomic alterations of key factors in NAFLD and fibrosis progression, we included 14 different transcriptomic datasets of liver tissues. In addition, two single-cell RNA sequencing (scRNA-seq) datasets were included to construct transcriptomic signatures that could represent specific cells. To explore the molecular subsets of fibrosis in NAFLD based on the transcriptomic features, we used a high-quality RNA-sequencing (RNA-seq) dataset of liver tissues from patients with NAFLD. Non-negative matrix factorization (NMF) was used to analyze the molecular subsets of NAFLD based on the gene set variation analysis (GSVA) enrichment scores of key molecule features in liver tissues.

RESULTS

The key transcriptomic signatures on NAFLD including non-alcoholic steatohepatitis (NASH) signature, fibrosis signature, non-alcoholic fatty liver (NAFL) signature, liver aging signature and TGF-β signature were constructed by liver transcriptome datasets. We analyzed two liver scRNA-seq datasets and constructed cell type-specific transcriptomic signatures based on the genes that were highly expressed in each cell subset. We analyzed the molecular subsets of NAFLD by NMF and categorized four main subsets of NAFLD. Cluster 4 subset is mainly characterized by liver fibrosis. Patients with Cluster 4 subset have more advanced liver fibrosis than patients with other subsets, or may have a high risk of liver fibrosis progression. Furthermore, we identified two key monocyte-macrophage subsets which were both significantly correlated with the progression of liver fibrosis in NAFLD patients.

CONCLUSION

Our study revealed the molecular subtypes of NAFLD by integrating key information from transcriptomic expression profiling and liver microenvironment, and identified a novel and distinct fibrosis subset of NAFLD. The fibrosis subset is significantly correlated with the profibrotic macrophages and M2 macrophage subset. These two liver macrophage subsets may be important players in the progression of liver fibrosis of NAFLD patients.

Missing data in multi-omics integration: Recent advances through artificial intelligence

Biological systems function through complex interactions between various 'omics (biomolecules), and a more complete understanding of these systems is only possible through an integrated, multi-omic perspective. This has presented the need for the development of integration approaches that are able to capture the complex, often non-linear, interactions that define these biological systems and are adapted to the challenges of combining the heterogenous data across 'omic views. A principal challenge to multi-omic integration is missing data because all biomolecules are not measured in all samples. Due to either cost, instrument sensitivity, or other experimental factors, data for a biological sample may be missing for one or more 'omic techologies. Recent methodological developments in artificial intelligence and statistical learning have greatly facilitated the analyses of multi-omics data, however many of these techniques assume access to completely observed data. A subset of these methods incorporate mechanisms for handling partially observed samples, and these methods are the focus of this review. We describe recently developed approaches, noting their primary use cases and highlighting each method's approach to handling missing data. We additionally provide an overview of the more traditional missing data workflows and their limitations; and we discuss potential avenues for further developments as well as how the missing data issue and its current solutions may generalize beyond the multi-omics context.

Heterogeneous data integration methods for patient similarity networks

Patient similarity networks (PSNs), where patients are represented as nodes and their similarities as weighted edges, are being increasingly used in clinical research. These networks provide an insightful summary of the relationships among patients and can be exploited by inductive or transductive learning algorithms for the prediction of patient outcome, phenotype and disease risk. PSNs can also be easily visualized, thus offering a natural way to inspect complex heterogeneous patient data and providing some level of explainability of the predictions obtained by machine learning algorithms. The advent of high-throughput technologies, enabling us to acquire high-dimensional views of the same patients (e.g. omics data, laboratory data, imaging data), calls for the development of data fusion techniques for PSNs in order to leverage this rich heterogeneous information. In this article, we review existing methods for integrating multiple biomedical data views to construct PSNs, together with the different patient similarity measures that have been proposed. We also review methods that have appeared in the machine learning literature but have not yet been applied to PSNs, thus providing a resource to navigate the vast machine learning literature existing on this topic. In particular, we focus on methods that could be used to integrate very heterogeneous datasets, including multi-omics data as well as data derived from clinical information and medical imaging.

Protocol for integrative subtyping of lower-grade gliomas using the SUMO pipeline

Grouping patients into subtypes with homogeneous molecular features can guide diagnosis and therapeutic interventions. SUMO is a computational pipeline that uses nonnegative matrix factorization of patient-similarity networks to integrate continuous multi-omic datasets for molecular subtyping of a disease. Here, we present a detailed protocol to demonstrate its use in determining subtypes of lower-grade gliomas by integrating gene expression, DNA methylation, and miRNA expression data from the TCGA-LGG cohort. For complete details on the use and execution of this profile, please refer to Sienkiewicz et al. (2022).