scGREAT: Transformer-based deep-language model for gene regulatory network inference from single-cell transcriptomics

Gene regulatory networks (GRNs) involve complex and multi-layer regulatory interactions between regulators and their target genes. Precise knowledge of GRNs is important in understanding cellular processes and molecular functions. Recent breakthroughs in single-cell sequencing technology made it possible to infer GRNs at single-cell level. Existing methods, however, are limited by expensive computations, and sometimes simplistic assumptions. To overcome these obstacles, we propose scGREAT, a framework to infer GRN using gene embeddings and transformer from single-cell transcriptomics. scGREAT starts by constructing gene expression and gene biotext dictionaries from scRNA-seq data and gene text information. The representation of TF gene pairs is learned through optimizing embedding space by transformer-based engine. Results illustrated scGREAT outperformed other contemporary methods on benchmarks. Besides, gene representations from scGREAT provide valuable gene regulation insights, and external validation on spatial transcriptomics illuminated the mechanism behind scGREAT annotation. Moreover, scGREAT identified several TF target regulations corroborated in studies.

ProteoMixture: A cell type deconvolution tool for bulk tissue proteomic data

Numerous multi-omic investigations of cancer tissue have documented varying and poor pairwise transcript:protein quantitative correlations, and most deconvolution tools aiming to predict cell type proportions (cell admixture) have been developed and credentialed using transcript-level data alone. To estimate cell admixture using protein abundance data, we analyzed proteome and transcriptome data generated from contrived admixtures of tumor, stroma, and immune cell models or those selectively harvested from the tissue microenvironment by laser microdissection from high grade serous ovarian cancer (HGSOC) tumors. Co-quantified transcripts and proteins performed similarly to estimate stroma and immune cell admixture (r ≥ 0.63) in two commonly used deconvolution algorithms, ESTIMATE or ConsensusTME. We further developed and optimized protein-based signatures estimating cell admixture proportions and benchmarked these using bulk tumor proteomic data from over 150 patients with HGSOC. The optimized protein signatures supporting cell type proportion estimates from bulk tissue proteomic data are available at https://lmdomics.org/ProteoMixture/.

Safeguarding authenticity for mitigating the harms of generative AI: Issues, research agenda, and policies for detection, fact-checking, and ethical AI

As the influence of transformer-based approaches in general and generative artificial intelligence (AI) in particular continues to expand across various domains, concerns regarding authenticity and explainability are on the rise. Here, we share our perspective on the necessity of implementing effective detection, verification, and explainability mechanisms to counteract the potential harms arising from the proliferation of AI-generated inauthentic content and science. We recognize the transformative potential of generative AI, exemplified by ChatGPT, in the scientific landscape. However, we also emphasize the urgency of addressing associated challenges, particularly in light of the risks posed by disinformation, misinformation, and unreproducible science. This perspective serves as a response to the call for concerted efforts to safeguard the authenticity of information in the age of AI. By prioritizing detection, fact-checking, and explainability policies, we aim to foster a climate of trust, uphold ethical standards, and harness the full potential of AI for the betterment of science and society.

In silico designing and characterization of outer membrane protein K (OmpK) from Vibrio anguillarum and its expression in Nicotiana tabacum for the development of a plant-based vaccine against fish vibriosis

Vibriosis is caused by Vibrio anguillarum in various species of aquaculture. A novel, secure, and stable vaccine is needed to eradicate vibriosis. Here, for reverse vaccinology and plant-based expression, the outer membrane protein K (OmpK) of V. anguillarum was chosen due to its conserved nature in all Vibrio species. OmpK, an ideal vaccine candidate against vibriosis, demonstrated immunogenic, non-allergic, and non-toxic behavior by using various bioinformatics tools. Docking showed the interaction of the OmpK model with TLR-5. In comparison to costly platforms, plants can be used as alternative and economic bio-factories to produce vaccine antigens. We expressed OmpK antigen in Nicotiana tabacum using Agrobacterium-mediated transformation. The expression vector was constructed using Gateway® cloning. Transgene integration was verified by polymerase chain reaction (PCR), and the copy number via qRT-PCR, which showed two copies of transgenes. Western blotting detected monomeric form of OmpK protein. The total soluble protein (TSP) fraction of OmpK was equivalent to 0.38% as detected by ELISA. Mice and fish were immunized with plant-derived OmpK antigen, which showed a significantly high level of anti-OmpK antibodies. The present study is the first report of OmpK antigen expression in higher plants for the potential use as vaccine in aquaculture against vibriosis, which could provide protection against multiple Vibrio species due to the conserved nature OmpK antigen.

FOTF-CPI: A compound-protein interaction prediction transformer based on the fusion of optimal transport fragments

Compound-protein interaction (CPI) affinity prediction plays an important role in reducing the cost and time of drug discovery. However, the interpretability of how fragments function in CPI is impacted by the fact that current methods ignore the affinity relationships between fragments of compounds and fragments of proteins in CPI modeling. This article introduces an improved Transformer called FOTF-CPI (a Fusion of Optimal Transport Fragments compound-protein interaction prediction model). We use an optimal transport-based fragmentation approach to improve the model's understanding of compound and protein sequences. Additionally, a fused attention mechanism is employed, which combines the features of fragments to capture full affinity information. This fused attention redistributes higher attention scores to fragments with higher affinity. Experimental results show FOTF-CPI achieves an average 2% higher performance than other models on all three datasets. Furthermore, the visualization confirms the potential of FOTF-CPI for drug discovery applications.

HLAEquity: Examining biases in pan-allele peptide-HLA binding predictors

Peptide-HLA (pHLA) binding prediction is essential in screening peptide candidates for personalized peptide vaccines. Machine learning (ML) pHLA binding prediction tools are trained on vast amounts of data and are effective in screening peptide candidates. Most ML models report the ability to generalize to HLA alleles unseen during training ("pan-allele" models). However, the use of datasets with imbalanced allele content raises concerns about biased model performance. First, we examine the data bias of two ML-based pan-allele pHLA binding predictors. We find that the pHLA datasets overrepresent alleles from geographic populations of high-income countries. Second, we show that the identified data bias is perpetuated within ML models, leading to algorithmic bias and subpar performance for alleles expressed in low-income geographic populations. We draw attention to the potential therapeutic consequences of this bias, and we challenge the use of the term "pan-allele" to describe models trained with currently available public datasets.

Gene-metabolite annotation with shortest reactional distance enhances metabolite genome-wide association studies results

Metabolite genome-wide association studies (mGWAS) have advanced our understanding of the genetic control of metabolite levels. However, interpreting these associations remains challenging due to a lack of tools to annotate gene-metabolite pairs beyond the use of conservative statistical significance threshold. Here, we introduce the shortest reactional distance (SRD) metric, drawing from the comprehensive KEGG database, to enhance the biological interpretation of mGWAS results. We applied this approach to three independent mGWAS, including a case study on sickle cell disease patients. Our analysis reveals an enrichment of small SRD values in reported mGWAS pairs, with SRD values significantly correlating with mGWAS p values, even beyond the standard conservative thresholds. We demonstrate the utility of SRD annotation in identifying potential false negatives and inaccuracies within current metabolic pathway databases. Our findings highlight the SRD metric as an objective, quantitative and easy-to-compute annotation for gene-metabolite pairs, suitable to integrate statistical evidence to biological networks.

Preliminary study of all-solid-state batteries: Evaluation of blast formation during the thermal runaway

All-solid-state batteries have been developed to increase energy density by replacing the lithiated graphite negative electrode by a lithium metal foil and to increase safety by removing the organic compounds. However, the safety issues of these batteries have received little attention up to now. The behavior of a reassembled all-solid-state battery under thermal stress was recorded by X-ray radiography and a high-speed camera. The thermal runaway (TR) lasted about 5 ms, thus extremely fast reaction kinetics. In comparison, the TR of a lithium-ion battery is about 500 ms. Furthermore, a 188-mbar aerial overpressure was measured using a piezoelectric sensor. Although this cell is not an explosive, 2.7 g TNT equivalent was calculated for it. This atypical behavior could have an impact on the casing or the battery pack. Therefore, it must be studied in greater detail.

A data-driven computational model for obesity-driven diabetes onset and remission through weight loss

Obesity is a major risk factor for the development of type 2 diabetes (T2D), where a sustained weight loss may result in T2D remission in individuals with obesity. To design effective and feasible intervention strategies to prevent or reverse T2D, it is imperative to study the progression of T2D and remission together. Unfortunately, this is not possible through experimental and observational studies. To address this issue, we introduce a data-driven computational model and use human data to investigate the progression of T2D with obesity and remission through weight loss on the same timeline. We identify thresholds for the emergence of T2D and necessary conditions for remission. We explain why remission is only possible within a window of opportunity and the way that window depends on the progression history of T2D, individual's metabolic state, and calorie restrictions. These findings can help to optimize therapeutic intervention strategies for T2D prevention or treatment.

A multi-center performance assessment for automated histopathological classification and grading of glioma using whole slide images

Accurate pathological classification and grading of gliomas is crucial in clinical diagnosis and treatment. The application of deep learning techniques holds promise for automated histological pathology diagnosis. In this study, we collected 733 whole slide images from four medical centers, of which 456 were used for model training, 150 for internal validation, and 127 for multi-center testing. The study includes 5 types of common gliomas. A subtask-guided multi-instance learning image-to-label training pipeline was employed. The pipeline leveraged "patch prompting" for the model to converge with reasonable computational cost. Experiments showed that an overall accuracy of 0.79 in the internal validation dataset. The performance on the multi-center testing dataset showed an overall accuracy to 0.73. The findings suggest a minor yet acceptable performance decrease in multi-center data, demonstrating the model's strong generalizability and establishing a robust foundation for future clinical applications.

Review: Computational analysis of human skeletal remains in ancient DNA and forensic genetics

Degraded DNA is used to answer questions in the fields of ancient DNA (aDNA) and forensic genetics. While aDNA studies typically center around human evolution and past history, and forensic genetics is often more concerned with identifying a specific individual, scientists in both fields face similar challenges. The overlap in source material has prompted periodic discussions and studies on the advantages of collaboration between fields toward mutually beneficial methodological advancements. However, most have been centered around wet laboratory methods (sampling, DNA extraction, library preparation, etc.). In this review, we focus on the computational side of the analytical workflow. We discuss limitations and considerations to consider when working with degraded DNA. We hope this review provides a framework to researchers new to computational workflows for how to think about analyzing highly degraded DNA and prompts an increase of collaboration between the forensic genetics and aDNA fields.

Genetic overlap for ten cardiovascular diseases: A comprehensive gene-centric pleiotropic association analysis and Mendelian randomization study

Recent studies suggest that pleiotropic effects may explain the genetic architecture of cardiovascular diseases (CVDs). We conducted a comprehensive gene-centric pleiotropic association analysis for ten CVDs using genome-wide association study (GWAS) summary statistics to identify pleiotropic genes and pathways that may underlie multiple CVDs. We found shared genetic mechanisms underlying the pathophysiology of CVDs, with over two-thirds of the diseases exhibiting common genes and single-nucleotide polymorphisms (SNPs). Significant positive genetic correlations were observed in more than half of paired CVDs. Additionally, we investigated the pleiotropic genes shared between different CVDs, as well as their functional pathways and distribution in different tissues. Moreover, six hub genes, including ALDH2, XPO1, HSPA1L, ESR2, WDR12, and RAB1A, as well as 26 targeted potential drugs, were identified. Our study provides further evidence for the pleiotropic effects of genetic variants on CVDs and highlights the importance of considering pleiotropy in genetic association studies.

An interpretable machine learning pipeline based on transcriptomics predicts phenotypes of lupus patients

Machine learning (ML) has the potential to identify subsets of patients with distinct phenotypes from gene expression data. However, phenotype prediction using ML has often relied on identifying important genes without a systems biology context. To address this, we created an interpretable ML approach based on blood transcriptomics to predict phenotype in systemic lupus erythematosus (SLE), a heterogeneous autoimmune disease. We employed a sequential grouped feature importance algorithm to assess the performance of gene sets, including immune and metabolic pathways and cell types, known to be abnormal in SLE in predicting disease activity and organ involvement. Gene sets related to interferon, tumor necrosis factor, the mitoribosome, and T cell activation were the best predictors of phenotype with excellent performance. These results suggest potential relationships between the molecular pathways identified in each model and manifestations of SLE. This ML approach to phenotype prediction can be applied to other diseases and tissues.

Prediction of postpartum prediabetes by machine learning methods in women with gestational diabetes mellitus

Early onset of type 2 diabetes and cardiovascular disease are common complications for women diagnosed with gestational diabetes. Prediabetes refers to a condition in which blood glucose levels are higher than normal, but not yet high enough to be diagnosed as type 2 diabetes. Currently, there is no accurate way of knowing which women with gestational diabetes are likely to develop postpartum prediabetes. This study aims to predict the risk of postpartum prediabetes in women diagnosed with gestational diabetes. Our sparse logistic regression approach selects only two variables - antenatal fasting glucose at OGTT and HbA1c soon after the diagnosis of GDM - as relevant, but gives an area under the receiver operating characteristic curve of 0.72, outperforming all other methods. We envision this to be a practical solution, which coupled with a targeted follow-up of high-risk women, could yield better cardiometabolic outcomes in women with a history of GDM.

Bayesian multivariate genetic analysis improves translational insights

While lipid traits are known essential mediators of cardiovascular disease, few approaches have taken advantage of their shared genetic effects. We apply a Bayesian multivariate size estimator, mash, to GWAS of four lipid traits in the Million Veterans Program (MVP) and provide posterior mean and local false sign rates for all effects. These estimates borrow information across traits to improve effect size accuracy. We show that controlling local false sign rates accurately and powerfully identifies replicable genetic associations and that multivariate control furthers the ability to explain complex diseases. Our application yields high concordance between independent datasets, more accurately prioritizes causal genes, and significantly improves polygenic prediction beyond state-of-the-art methods by up to 59% for lipid traits. The use of Bayesian multivariate genetic shrinkage has yet to be applied to human quantitative trait GWAS results, and we present a staged approach to prediction on a polygenic scale.

Gene presence/absence variation in Mytilus galloprovincialis and its implications in gene expression and adaptation

Presence/absence variation (PAV) is a well-known phenomenon in prokaryotes that was described for the first time in bivalves in 2020 in Mytilus galloprovincialis. The objective of the present study was to further our understanding of the PAV phenomenon in mussel biology. The distribution of PAV was studied in a mussel chromosome-level genome assembly, revealing a widespread distribution but with hotspots of dispensability. Special attention was given to the effect of PAV in gene expression, since dispensable genes were found to be inherently subject to distortions due to their sparse distribution among individuals. Furthermore, the high expression and strong tissue specificity of some dispensable genes, such as myticins, strongly supported their biological relevance. The significant differences in the repertoire of dispensable genes associated with two geographically distinct populations suggest that PAV is involved in local adaptation. Overall, the PAV phenomenon would provide a key selective advantage at the population level.

Systematic analysis reveals a pan-cancer SNHG family signature predicting prognosis and immunotherapy response

Small nucleolar RNA host genes (SNHGs) are a special family of long non-coding RNAs (lncRNAs), which not only function in a way similar to other lncRNAs but also influence the intracellular level of small nucleolar RNAs to modulate cancers. However, the features of SNHGs and their role in the prognosis and immunotherapeutic response of human cancer have not been explored. We found that SNHGs were commonly deregulated and correlated with patient survival in various cancers. The critical role of DNA methylation and somatic alterations on deregulation was also identified. SNHG family score was significantly associated with survival, multiple tumor characteristics, and tumor microenvironment. SNHG-related risk score could serve as a prognostic and immunotherapeutic response biomarker based on multiple databases. This study emphasizes the potential of SNHGs as biomarkers for prognosis and immunotherapeutic response, enabling further research into the immune regulatory mechanism and therapeutic potentials of SNHGs in cancer.

A physiologically based model of bile acid metabolism in mice

Bile acid (BA) metabolism is a complex system that includes a wide variety of primary and secondary, as well as conjugated and unconjugated BAs that undergo continuous enterohepatic circulation (EHC). Alterations in both composition and dynamics of BAs have been associated with various diseases. However, a mechanistic understanding of the relationship between altered BA metabolism and related diseases is lacking. Computational modeling may support functional analyses of the physiological processes involved in the EHC of BAs along the gut-liver axis. In this study, we developed a physiologically based model of murine BA metabolism describing synthesis, hepatic and microbial transformations, systemic distribution, excretion, and EHC of BAs at the whole-body level. For model development, BA metabolism of specific pathogen-free (SPF) mice was characterized in vivo by measuring BA levels and composition in various organs, expression of transporters along the gut, and cecal microbiota composition. We found significantly different BA levels between male and female mice that could only be explained by adjusted expression of the hepatic enzymes and transporters in the model. Of note, this finding was in agreement with experimental observations. The model for SPF mice could also describe equivalent experimental data in germ-free mice by specifically switching off microbial activity in the intestine. The here presented model can therefore facilitate and guide functional analyses of BA metabolism in mice, e.g., the effect of pathophysiological alterations on BA metabolism and translation of results from mouse studies to a clinically relevant context through cross-species extrapolation.

Radiomics and deep learning models to differentiate lung adenosquamous carcinoma: A multicenter trial

Adenosquamous carcinoma (ASC) is frequently misdiagnosed or overlooked in clinical practice due to its dual histological components and potential transformation from either adenocarcinoma (ADC) or squamous cell carcinoma (SCC). Our study aimed to differentiate ASC from ADC and SCC by incorporating features of enhanced CTs and clinical characteristics to build radiomics and deep learning models. The classification models were trained in Xiangya Hospital and validated in two other independent hospitals. The areas under the receiver operating characteristic curves (AUC), accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were used to estimate the performance. The optimal three-class classification model achieved a maximum AUC of 0.89 and accuracy of 0.81 in external validation sets, AUC of 0.99 and accuracy of 0.99 in the internal test set. These findings highlight the efficacy of our models in differentiating ASC, providing a non-invasive, timely, and accurate diagnostic approach before and during the treatment.

A personalized stepwise dynamic predictive algorithm of the time to first treatment in chronic lymphocytic leukemia

Personalized prediction is ideal in chronic lymphocytic leukemia (CLL). Although refined models have been developed, stratifying patients in risk groups, it is required to accommodate time-dependent information of patients, to address the clinical heterogeneity observed within these groups. In this direction, this study proposes a personalized stepwise dynamic predictive algorithm (PSDPA) for the time-to-first-treatment of the individual patient. The PSDPA introduces a personalized Score, reflecting the evolution in the patient's follow-up, employed to develop a reference pool of patients. Score evolution's similarity is used to predict, at a selected time point, the time-to-first-treatment for a new patient. Additional patient's biological information may be utilized. The algorithm was applied to 20 CLL patients, indicating that stricter assessment criteria for the Score evolution's similarity, and biological similarity exploitation, may improve prediction. The PSDPA capitalizes on both the follow-up and the biological background of the individual patient, dynamically promoting personalized prediction in CLL.

Atherosclerotic plaque vulnerability quantification system for clinical and biological interpretability

Acute myocardial infarction dominates coronary artery disease mortality. Identifying bio-signatures for plaque destabilization and rupture is important for preventing the transition from coronary stability to instability and the occurrence of thrombosis events. This computational systems biology study enrolled 2,235 samples from 22 independent bulks cohorts and 14 samples from two single-cell cohorts. A machine-learning integrative program containing nine learners was developed to generate a warning classifier linked to atherosclerotic plaque vulnerability signature (APVS). The classifier displays the reliable performance and robustness for distinguishing ST-elevation myocardial infarction from chronic coronary syndrome at presentation, and revealed higher accuracy to 33 pathogenic biomarkers. We also developed an APVS-based quantification system (APVSLevel) for comprehensively quantifying atherosclerotic plaque vulnerability, empowering early-warning capabilities, and accurate assessment of atherosclerosis severity. It unraveled the multidimensional dysregulated mechanisms at high resolution. This study provides a potential tool for macro-level differential diagnosis and evaluation of subtle genetic pathological changes in atherosclerosis.

Understanding locomotion in trilobites by means of three-dimensional models

Trilobites were one of the first animals on Earth to leave their imprints on the seafloor. Such imprints represent behavioral traces related to feeding or protection, in both cases implying different types of locomotion. Modeling how trilobites moved is essential to understand their evolutionary history and ecological impact on marine substrates. Herein, locomotion in trilobites is approached by means of three-dimensional models, which yielded two main gait types. These two gaits reflect basic behaviors: burrowing and walking. This model reveals that trilobites could change their gait and consequently increase rapidly their speed varying the amplitude of the metachronal wave, a change independent from their biological structure. Fast increases in speed enhanced the protection of trilobites against predators and sudden environmental crises. The trilobite body pattern constrained their gaits, controlled by the distance between the pair of legs and between legs in a same segment.

Selection for immune evasion in SARS-CoV-2 revealed by high-resolution epitope mapping and sequence analysis

Here, we exploit a deep serological profiling strategy coupled with an integrated, computational framework for the analysis of SARS-CoV-2 humoral immune responses. Applying a high-density peptide array (HDPA) spanning the entire proteomes of SARS-CoV-2 and endemic human coronaviruses allowed identification of B cell epitopes and relate them to their evolutionary and structural properties. We identify hotspots of pre-existing immunity and identify cross-reactive epitopes that contribute to increasing the overall humoral immune response to SARS-CoV-2. Using a public dataset of over 38,000 viral genomes from the early phase of the pandemic, capturing both inter- and within-host genetic viral diversity, we determined the evolutionary profile of epitopes and the differences across proteins, waves, and SARS-CoV-2 variants. Lastly, we show that mutations in spike and nucleocapsid epitopes are under stronger selection between than within patients, suggesting that most of the selective pressure for immune evasion occurs upon transmission between hosts.

Application of a deep learning algorithm in the detection of hip fractures

This paper describes the development of a deep learning model for prediction of hip fractures on pelvic radiographs (X-rays). Developed using over 40,000 pelvic radiographs from a single institution, the model demonstrated high sensitivity and specificity when applied to a test set of emergency department radiographs. This study approximates the real-world application of a deep learning fracture detection model by including radiographs with sub-optimal image quality, other non-hip fractures, and metallic implants, which were excluded from prior published work. The study also explores the effect of ethnicity on model performance, as well as the accuracy of visualization algorithm for fracture localization.

Intragenomic diversity of the V9 hypervariable domain in eukaryotes has little effect on metabarcoding

Metabarcoding revolutionized our understanding of diversity and ecology of microorganisms in different habitats. However, it is also associated with several inherent biases, one of which is associated with intragenomic diversity of a molecular barcode. Here, we compare intragenomic variability of the V9 region of the 18S rRNA gene in 19 eukaryotic phyla abundant in marine plankton. The level of intragenomic variability is comparable across all the phyla, and in most genomes and transcriptomes one V9 sequence and one OTU is predominant. However, most of the variability observed at the barcode level is probably caused by sequencing errors and is mitigated by using a denoising tool, DADA2. The SWARM algorithm commonly used in metabarcoding studies is not optimal for collapsing genuine and erroneous sequences into a single OTU, leading to an overestimation of diversity in metabarcoding data. For an unknown reason, SWARM inflates diversity of eupelagonemids more than that of other eukaryotes.

Multi-scale spatial modeling of immune cell distributions enables survival prediction in primary central nervous system lymphoma

To understand the clinical significance of the tumor microenvironment (TME), it is essential to study the interactions between malignant and non-malignant cells in clinical specimens. Here, we established a computational framework for a multiplex imaging system to comprehensively characterize spatial contexts of the TME at multiple scales, including close and long-distance spatial interactions between cell type pairs. We applied this framework to a total of 1,393 multiplex imaging data newly generated from 88 primary central nervous system lymphomas with complete follow-up data and identified significant prognostic subgroups mainly shaped by the spatial context. A supervised analysis confirmed a significant contribution of spatial context in predicting patient survival. In particular, we found an opposite prognostic value of macrophage infiltration depending on its proximity to specific cell types. Altogether, we provide a comprehensive framework to analyze spatial cellular interaction that can be broadly applied to other technologies and tumor contexts.

A contamination focused approach for optimizing the single-cell RNA-seq experiment

Droplet-based single-cell RNA-seq (scRNA-seq) data are plagued by ambient contaminations caused by nucleic acid material released by dead and dying cells. This material is mixed into the buffer and is co-encapsulated with cells, leading to a lower signal-to-noise ratio. Although there exist computational methods to remove ambient contaminations post-hoc, the reliability of algorithms in generating high-quality data from low-quality sources remains uncertain. Here, we assess data quality before data filtering by a set of quantitative, contamination-based metrics that assess data quality more effectively than standard metrics. Through a series of controlled experiments, we report improvements that can minimize ambient contamination outside of tissue dissociation, via cell fixation, improved cell loading, microfluidic dilution, and nuclei versus cell preparation; many of these parameters are inaccessible on commercial platforms. We provide end-users with insights on factors that can guide their decision-making regarding optimizations that minimize ambient contamination, and metrics to assess data quality.

Whole-genome DNA methylation and DNA methylation-based biomarkers in lung squamous cell carcinoma

Exploring early detection methods through comprehensive evaluation of DNA methylation for lung squamous cell carcinoma (LUSC) patients is of great significance. By using different machine learning algorithms for feature selection and model construction based on The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases, five methylation biomarkers in LUSC (along with mapped genes) were identified including cg14823851 (TBX4), cg02772121 (TRIM15), cg10424681 (C6orf201), cg12910906 (ARHGEF4), and cg20181079 (OR4D11), achieving extremely high sensitivity and specificity in distinguishing LUSC from normal samples in independent cohorts. Pyrosequencing assay verified DNA methylation levels, meanwhile qRT-PCR and immunohistochemistry results presented their accordant methylation-related gene expression statuses in paired LUSC and normal lung tissues. The five methylation-based biomarkers proposed in this study have great potential for the diagnosis of LUSC and could guide studies in methylation-regulated tumor development and progression.

An agent-based model of monocyte differentiation into tumour-associated macrophages in chronic lymphocytic leukemia

Monocyte-derived macrophages help maintain tissue homeostasis and defend the organism against pathogens. In tumors, recent studies have uncovered complex macrophage populations, including tumor-associated macrophages, which support tumorigenesis through cancer hallmarks such as immunosuppression, angiogenesis, or matrix remodeling. In the case of chronic lymphocytic leukemia, these macrophages are known as nurse-like cells (NLCs) and they protect leukemic cells from spontaneous apoptosis, contributing to their chemoresistance. We propose an agent-based model of monocyte differentiation into NLCs upon contact with leukemic B cells in vitro. We performed patient-specific model optimization using cultures of peripheral blood mononuclear cells from patients. Using our model, we were able to reproduce the temporal survival dynamics of cancer cells in a patient-specific manner and to identify patient groups related to distinct macrophage phenotypes. Our results show a potentially important role of phagocytosis in the polarization process of NLCs and in promoting cancer cells' enhanced survival.

Dissecting molecular, pathological, and clinical features associated with tumor neural/neuroendocrine heterogeneity

Lineage plasticity, especially transdifferentiation between neural/neuroendocrine (NE) and non-NE lineage, has been observed in multiple cancer types and linked to increased tumor aggressiveness. However, existing NE/non-NE subtype classifications in various cancer types were established through ad hoc approaches in different studies, making it difficult to align findings across cancer types and extend investigations to new datasets. To address this issue, we developed a generalized strategy to generate quantitative NE scores and a web application to facilitate its implementation. We applied this method to nine datasets covering seven cancer types, including two neural cancers, two neuroendocrine cancers, and three non-NE cancers. Our analysis revealed significant NE inter-tumoral heterogeneity and identified strong associations between NE scores and molecular, histological, and clinical features, including prognosis in different cancer types. These results support the translational utility of NE scores. Overall, our work demonstrated a broadly applicable strategy for determining the NE properties of tumors.

Visualizing gene expression changes in time, space, and single cells with expressyouRcell

The last decade has witnessed massive advancements in high-throughput techniques capable of producing increasingly complex gene expression datasets across time and space and at the resolution of single cells. Yet, the large volume of big data available and the complexity of experimental designs hamper an easy understanding and effective communication of the results. We present expressyouRcell, an easy-to-use R package to map the multi-dimensional variations of transcript and protein levels in dynamic cell pictographs. expressyouRcell visualizes gene expression variations as pictographic representations of cell-type thematic maps. expressyouRcell visually reduces the complexity of displaying gene expression and protein level changes across multiple measurements (time points or single-cell trajectories) by generating dynamic representations of cellular pictographs. We applied expressyouRcell to single cell, bulk RNA sequencing (RNA-seq), and proteomics datasets, demonstrating its flexibility and usability in the visualization of complex variations in gene expression. Our approach improves the standard quantitative interpretation and communication of relevant results.

Modeling breast cancer proliferation, drug synergies, and alternating therapies

Estrogen receptor positive (ER+) breast cancer is responsive to a number of targeted therapies used clinically. Unfortunately, the continuous application of targeted therapy often results in resistance, driving the consideration of combination and alternating therapies. Toward this end, we developed a mathematical model that can simulate various mono, combination, and alternating therapies for ER + breast cancer cells at different doses over long time scales. The model is used to look for optimal drug combinations and predicts a significant synergism between Cdk4/6 inhibitors in combination with the anti-estrogen fulvestrant, which may help explain the clinical success of adding Cdk4/6 inhibitors to anti-estrogen therapy. Furthermore, the model is used to optimize an alternating treatment protocol so it works as well as monotherapy while using less total drug dose.

An artificial bee bare-bone hunger games search for global optimization and high-dimensional feature selection

The domains of contemporary medicine and biology have generated substantial high-dimensional genetic data. Identifying representative genes and decreasing the dimensionality of the data can be challenging. The goal of gene selection is to minimize computing costs and enhance classification precision. Therefore, this article designs a new wrapper gene selection algorithm named artificial bee bare-bone hunger games search (ABHGS), which is the hunger games search (HGS) integrated with an artificial bee strategy and a Gaussian bare-bone structure to address this issue. To evaluate and validate the performance of our proposed method, ABHGS is compared to HGS and a single strategy embedded in HGS, six classic algorithms, and ten advanced algorithms on the CEC 2017 functions. The experimental results demonstrate that the bABHGS outperforms the original HGS. Compared to peers, it increases classification accuracy and decreases the number of selected features, indicating its actual engineering utility in spatial search and feature selection.

Topologically associating domain underlies tissue specific expression of long intergenic non-coding RNAs

Accumulating evidence indicates that long intergenic non-coding RNAs (lincRNAs) show more tissue-specific expression patterns than protein-coding genes (PCGs). However, although lincRNAs are subject to canonical transcriptional regulation like PCGs, the molecular basis for the specificity of their expression patterns remains unclear. Here, using expression data and coordinates of topologically associating domains (TADs) in human tissues, we show that lincRNA loci are significantly enriched in the more internal region of TADs compared to PCGs and that lincRNAs within TADs have higher tissue specificity than those outside TADs. Based on these, we propose an analytical framework to interpret transcriptional status using lincRNA as an indicator. We applied it to hypertrophic cardiomyopathy data and found disease-specific transcriptional regulation: ectopic expression of keratin at the TAD level and derepression of myocyte differentiation-related genes by E2F1 with down-regulation of LINC00881. Our results provide understanding of the function and regulation of lincRNAs according to genomic structure.

Pseudo-temporal dynamics of chemoresistant triple negative breast cancer cells reveal EGFR/HER2 inhibition as synthetic lethal during mid-neoadjuvant chemotherapy

In the absence of targetable hormonal axes, chemoresistance for triple-negative breast cancer (TNBC) often compromises patient outcomes. To investigate the underlying tumor dynamics, we performed trajectory analysis on the single-nuclei RNA-seq (snRNA-seq) of chemoresistant tumor clones during neoadjuvant chemotherapy (NAC). It revealed a common tumor trajectory across multiple patients with HER2-like expansions during NAC. Genome-wide CRISPR-Cas9 knock-out on mammary epithelial cells revealed chemosensitivity-promoting knock-outs were up-regulated along the tumor trajectory. Furthermore, we derived a consensus gene signature of TNBC chemoresistance by comparing the trajectory transcriptome with chemoresistant transcriptomes from TNBC cell lines and poor prognosis patient samples to predict FDA-approved drugs, including afatinib (pan-HER inhibitor), targeting the consensus signature. We validated the synergistic efficacy of afatinib and paclitaxel in chemoresistant TNBC cells and confirmed pharmacological suppression of the consensus signature. The study provides a dynamic model of chemoresistant tumor transcriptome, and computational framework for pharmacological intervention.

A Kmer-based paired-end read de novo assembler and genotyper for canine MHC class I genotyping

The major histocompatibility complex class I (MHC-I) genes are highly polymorphic. MHC-I genotyping is required for determining the peptide epitopes available to an individual's T-cell repertoire. Current genotyping software tools do not work for the dog, due to very limited known canine alleles. To address this, we developed a Kmer-based paired-end read (KPR) de novo assembler and genotyper, which assemble paired-end RNA-seq reads from MHC-I regions into contigs, and then genotype each contig and estimate its expression level. KPR tools outperform other popular software examined in typing new alleles. We used KPR tools to successfully genotype152 dogs from a published dataset. The study discovers 33 putative new alleles, finds dominant alleles in 4 dog breeds, and builds allele diversity and expression landscapes among the 152 dogs. Our software meets a significant need in biomedical research.

Quantitative analysis of disease-related metabolic dysregulation of human microbiota

The metabolic activity of all the micro-organism composing the human microbiome interacts with the host metabolism contributing to human health and disease in a way that is not fully understood. Here, we introduce STELLA, a computational method to derive the spectrum of metabolites associated with the microbiome of an individual. STELLA integrates known information on metabolic pathways associated with each bacterial species and extracts from these the list of metabolic products of each singular reaction by means of automatic text analysis. By comparing the result obtained on a single subject with the metabolic profile data of a control set of healthy subjects, we are able to identify individual metabolic alterations. To illustrate the method, we present applications to autism spectrum disorder and multiple sclerosis.

Polyvalent guide RNAs for CRISPR antivirals

CRISPR effector Cas13 recognizes and degrades RNA molecules that are complementary to its guide RNA (gRNA) and possesses potential as an antiviral biotechnology because it can degrade viral mRNA and RNA genomes. Because multiplexed targeting is a critical strategy to improve viral suppression, we developed a strategy to design of gRNAs where individual gRNAs have maximized activity at multiple viral targets, simultaneously, by exploiting the molecular biophysics of promiscuous target recognition by Cas13. These "polyvalent" gRNA sequences ("pgRNAs") provide superior antiviral elimination across tissue/organ scales in a higher organism (Nicotiana benthamiana) compared to conventionally-designed gRNAs-reducing detectable viral RNA by >30-fold, despite lacking perfect complementarity with either of their targets and, when multiplexed, reducing viral RNA by >99.5%. Pairs of pgRNA-targetable sequences are abundant in the genomes of RNA viruses, and this work highlights the need for specific approaches to the challenges of targeting viruses in eukaryotes using CRISPR.

Agent-based computational modeling of glioblastoma predicts that stromal density is central to oncolytic virus efficacy

Oncolytic viruses (OVs) are emerging cancer immunotherapy. Despite notable successes in the treatment of some tumors, OV therapy for central nervous system cancers has failed to show efficacy. We used an ex vivo tumor model developed from human glioblastoma tissue to evaluate the infiltration of herpes simplex OV rQNestin (oHSV-1) into glioblastoma tumors. We next leveraged our data to develop a computational, model of glioblastoma dynamics that accounts for cellular interactions within the tumor. Using our computational model, we found that low stromal density was highly predictive of oHSV-1 therapeutic success, suggesting that the efficacy of oHSV-1 in glioblastoma may be determined by stromal-to-tumor cell regional density. We validated these findings in heterogenous patient samples from brain metastatic adenocarcinoma. Our integrated modeling strategy can be applied to suggest mechanisms of therapeutic responses for central nervous system cancers and to facilitate the successful translation of OVs into the clinic.

Combined assessment of MHC binding and antigen abundance improves T cell epitope predictions

Many steps of the MHC class I antigen processing pathway can be predicted using computational methods. Here we show that epitope predictions can be further improved by considering abundance levels of peptides' source proteins. We utilized biophysical principles and existing MHC binding prediction tools in concert with abundance estimates of source proteins to derive a function that estimates the likelihood of a peptide to be an MHC class I ligand. We found that this combination improved predictions for both naturally eluted ligands and cancer neoantigen epitopes. We compared the use of different measures of antigen abundance, including mRNA expression by RNA-Seq, gene translation by Ribo-Seq, and protein abundance by proteomics on a dataset of SARS-CoV-2 epitopes. Epitope predictions were improved above binding predictions alone in all cases and gave the highest performance when using proteomic data. Our results highlight the value of incorporating antigen abundance levels to improve epitope predictions.

•

HLA ligands originate from highly expressed transcripts

•

Antigen abundance and HLA binding are independent predictors of ligands and epitopes

•

Utilizing RNA-Seq, Ribo-Seq, or proteomic data improves epitope predictions

•

Cancer-type-matched TCGA RNA-Seq data can be used to estimate gene expression in patient

Analysis of age-related left ventricular collagen remodeling in living donors: Implications in arrhythmogenesis

Age-related fibrosis in the left ventricle (LV) has been mainly studied in animals by assessing collagen content. Using second-harmonic generation microscopy and image processing, we evaluated amount, aggregation and spatial distribution of LV collagen in young to old pigs, and middle-age and elder living donors. All collagen features increased when comparing adult and old pigs with young ones, but not when comparing adult with old pigs or middle-age with elder individuals. Remarkably, all collagen parameters strongly correlated with lipofuscin, a biological age marker, in humans. By building patient-specific models of human ventricular tissue electrophysiology, we confirmed that amount and organization of fibrosis modulated arrhythmia vulnerability, and that distribution should be accounted for arrhythmia risk assessment. In conclusion, we characterize the age-associated changes in LV collagen and its potential implications for ventricular arrhythmia development. Consistency between pig and human results substantiate the pig as a relevant model of age-related LV collagen dynamics.

•

Collagen remodeling traits change from youth to adulthood, not from midlife onwards

•

In humans, collagen remodeling traits relate with the biological age-pigment lipofuscin

•

Beyond collagen amount, its distribution also influences ventricular arrhythmogenesis

•

Consistent age-related remodeling was observed amid healthy farm pigs and living donors

Quantifying the impact of physical activity on future glucose trends using machine learning

Prevention of hypoglycemia (glucose <70 mg/dL) during aerobic exercise is a major challenge in type 1 diabetes. Providing predictions of glycemic changes during and following exercise can help people with type 1 diabetes avoid hypoglycemia. A unique dataset representing 320 days and 50,000 + time points of glycemic measurements was collected in adults with type 1 diabetes who participated in a 4-arm crossover study evaluating insulin-pump therapies, whereby each participant performed eight identically designed in-clinic exercise studies. We demonstrate that even under highly controlled conditions, there is considerable intra-participant and inter-participant variability in glucose outcomes during and following exercise. Participants with higher aerobic fitness exhibited significantly lower minimum glucose and steeper glucose declines during exercise. Adaptive, personalized machine learning (ML) algorithms were designed to predict exercise-related glucose changes. These algorithms achieved high accuracy in predicting the minimum glucose and hypoglycemia during and following exercise sessions, for all fitness levels.

Deep learning model of somatic hypermutation reveals importance of sequence context beyond hotspot targeting

B cells undergo somatic hypermutation (SHM) of the Immunoglobulin (Ig) variable region to generate high-affinity antibodies. SHM relies on the activity of activation-induced deaminase (AID), which mutates C>U preferentially targeting WRC (W=A/T, R=A/G) hotspots. Downstream mutations at WA Polymerase η hotspots contribute further mutations. Computational models of SHM can describe the probability of mutations essential for vaccine responses. Previous studies using short subsequences (k-mers) failed to explain divergent mutability for the same k-mer. We developed the DeepSHM (Deep learning on SHM) model using k-mers of size 5-21, improving accuracy over previous models. Interpretation of DeepSHM identified an extended WWRCT motif with particularly high mutability. Increased mutability was further associated with lower surrounding G content. Our model also discovered a conserved AGYCTGGGGG (Y=C/T) motif within FW1 of IGHV3 family genes with unusually high T>G substitution rates. Thus, a wider sequence context increases predictive power and identifies features that drive mutational targeting.

Cross-species transcriptomic signatures predict response to MK2 inhibition in mouse models of chronic inflammation

Inflammatory bowel diseases (IBDs) are genetically complex and exhibit significant inter-patient heterogeneity in disease presentation and therapeutic response. Here, we show that mouse models of IBD exhibit variable responses to inhibition of MK2, a pro-inflammatory serine/threonine kinase, and that MK2 inhibition suppresses inflammation by targeting inflammatory monocytes and neutrophils in murine models. Using a computational approach (TransComp-R) that allows for cross-species comparison of transcriptomic features, we identified an IBD patient subgroup that is predicted to respond to MK2 inhibition, and an independent preclinical model of chronic intestinal inflammation predicted to be non-responsive, which we validated experimentally. Thus, cross-species mouse-human translation approaches can help to identify patient subpopulations in which to deploy new therapies.

•

MK2 kinase inhibition shows variable efficacy in different IBD mouse models

•

TCT and TNFΔARE mice express distinct inflammatory and MK2-responsive genes

•

“Response to MK2i” signature is enriched in monocytes and neutrophils

•

Cross-species modeling identifies patient groups potentially responsive to MK2i

TranSNPs: A class of functional SNPs affecting mRNA translation potential revealed by fraction-based allelic imbalance

Few studies have explored the association between SNPs and alterations in mRNA translation potential. We developed an approach to identify SNPs that can mark allele-specific protein expression levels and could represent sources of inter-individual variation in disease risk. Using MCF7 cells under different treatments, we performed polysomal profiling followed by RNA sequencing of total or polysome-associated mRNA fractions and designed a computational approach to identify SNPs showing a significant change in the allelic balance between total and polysomal mRNA fractions. We identified 147 SNPs, 39 of which located in UTRs. Allele-specific differences at the translation level were confirmed in transfected MCF7 cells by reporter assays. Exploiting breast cancer data from TCGA we identified UTR SNPs demonstrating distinct prognosis features and altering binding sites of RNA-binding proteins. Our approach produced a catalog of tranSNPs, a class of functional SNPs associated with allele-specific translation and potentially endowed with prognostic value for disease risk.

Automated design of CRISPR prime editors for 56,000 human pathogenic variants

Prime editors (PEs) are clustered regularly interspaced short palindromic repeats (CRISPR)-based genome engineering tools that can introduce precise base-pair edits. We developed an automated pipeline to correct (therapeutic editing) or introduce (disease modeling) human pathogenic variants from ClinVar that optimizes the design of several RNA constructs required for prime editing and avoids predicted off-targets in the human genome. However, using optimal PE design criteria, we find that only a small fraction of these pathogenic variants can be targeted. Through the use of alternative Cas9 enzymes and extended templates, we increase the number of targetable pathogenic variants from 32,000 to 56,000 variants and make these pre-designed PE constructs accessible through a web-based portal (http://primeedit.nygenome.org). Given the tremendous potential for therapeutic gene editing, we also assessed the possibility of developing universal PE constructs, finding that common genetic variants impact only a small minority of designed PEs.

•

CRISPR prime editors for therapeutic gene editing and disease modeling

•

Increase targetable variants with template extension and alternative Cas9 design

•

Optimize prime editors with off-target avoidance and integration of common variants

•

Web-tool for rapid prime editor design using gene name or ClinVar identifier

Characterization of mechanisms positioning costimulatory complexes in immune synapses

Small immunoglobulin superfamily (sIGSF) adhesion complexes form a corolla of microdomains around an integrin ring and secretory core during immunological synapse (IS) formation. The corolla recruits and retains major costimulatory/checkpoint complexes, such as CD28, making forces that govern corolla formation of particular interest. Here, we investigated the mechanisms underlying molecular reorganization of CD2, an adhesion and costimulatory molecule of the sIGSF family during IS formation. Computer simulations showed passive distal exclusion of CD2 complexes under weak interactions with the ramified F-actin transport network. Attractive forces between CD2 and CD28 complexes relocate CD28 from the IS center to the corolla. Size-based sorting interactions with large glycocalyx components, such as CD45, or short-range CD2 self-attraction successfully explain the corolla 'petals.' This establishes a general simulation framework for complex pattern formation observed in cell-bilayer and cell-cell interfaces, and the suggestion of new therapeutic targets, where boosting or impairing characteristic pattern formation can be pivotal.

MicroRNA function can be reversed by altering target gene expression levels

Paradoxically, many microRNAs appear to exhibit entirely opposite functions when placed in different contexts. For example, miR-125b has been shown to be pro-apoptotic in some studies, but anti-apoptotic in others. To investigate this phenomenon, we combine computational modeling with experimental approaches to examine how the function of miR-125b in apoptosis varies with respect to the expression levels of its pro-apoptotic and anti-apoptotic targets. In doing so, we elucidate a general trend that miR-125b is more pro-apoptotic when its anti-apoptotic targets are overexpressed, whereas it is more anti-apoptotic when its pro-apoptotic targets are overexpressed. We show that it is possible to completely reverse miR-125b′s function in apoptosis by modifying the expression levels of its target genes. Furthermore, miR-125b′s function may also be altered by the presence of anticancer drugs. These results suggest that the function of a microRNA can vary substantially and is dependent on its target gene expression levels.

•

Many miRNAs exhibit entirely opposite functions when placed in different contexts

•

miR-125b can be pro- or anti-apoptotic depending on target gene expression levels

•

The function of a miRNA can be reversed by altering target gene expression levels

•

The presence of anticancer drugs can also alter a miRNA's function

Prediction and identification of synergistic compound combinations against pancreatic cancer cells

Resistance to current therapies is common for pancreatic cancer and hence novel treatment options are urgently needed. In this work, we developed and validated a computational method to select synergistic compound combinations based on transcriptomic profiles from both the disease and compound side, combined with a pathway scoring system, which was then validated prospectively by testing 30 compounds (and their combinations) on PANC-1 cells. Some compounds selected as single agents showed lower GI50 values than the standard of care, gemcitabine. Compounds suggested as combination agents with standard therapy gemcitabine based on the best performing scoring system showed on average 2.82-5.18 times higher synergies compared to compounds that were predicted to be active as single agents. Examples of highly synergistic in vitro validated compound pairs include gemcitabine combined with Entinostat, thioridazine, loperamide, scriptaid and Saracatinib. Hence, the computational approach presented here was able to identify synergistic compound combinations against pancreatic cancer cells.

Activation of gene expression by detergent-like protein domains

The mechanisms by which transcriptional activation domains (tADs) initiate eukaryotic gene expression have been an enigma for decades because most tADs lack specificity in sequence, structure, and interactions with targets. Machine learning analysis of data sets of tAD sequences generated in vivo elucidated several functionality rules: the functional tAD sequences should (i) be devoid of or depleted with basic amino acid residues, (ii) be enriched with aromatic and acidic residues, (iii) be with aromatic residues localized mostly near the terminus of the sequence, and acidic residues localized more internally within a span of 20-30 amino acids, (iv) be with both aromatic and acidic residues preferably spread out in the sequence and not clustered, and (v) not be separated by occasional basic residues. These and other more subtle rules are not absolute, reflecting absence of a tAD consensus sequence, enormous variability, and consistent with surfactant-like tAD biochemical properties. The findings are compatible with the paradigm-shifting nucleosome detergent mechanism of gene expression activation, contributing to the development of the liquid-liquid phase separation model and the biochemistry of near-stochastic functional allosteric interactions.