MOTIVATION: Despite the success of recent machine learning algorithms' applications to survival analysis, their black-box nature hinders interpretability, which is arguably the most important aspect. Similarly, multi-omics data integration for survival analysis is often constrained by the underlying relationships and correlations that are rarely well understood. The goal of this work is to alleviate the interpretability problem in machine learning approaches for survival analysis and also demonstrate how multi-omics data integration improves survival analysis and pathway enrichment. We use meta-learning, a machine-learning algorithm that is trained on a variety of related datasets and allows quick adaptations to new tasks, to perform survival analysis and pathway enrichment on pan-cancer datasets. In recent machine learning research, meta-learning has been effectively used for knowledge transfer among multiple related datasets. RESULTS: We use meta-learning with Cox hazard loss to show that the integration of TCGA pan-cancer data increases the performance of survival analysis. We also apply advanced model interpretability method called DeepLIFT (Deep Learning Important FeaTures) to show different sets of enriched pathways for multi-omics and transcriptomics data. Our results show that multi-omics cancer survival analysis enhances performance compared with using transcriptomics or clinical data alone. Additionally, we show a correlation between variable importance assignment from DeepLIFT and gene coenrichment, suggesting that genes with higher and similar contribution scores are more likely to be enriched together in the same enrichment sets. AVAILABILITY AND IMPLEMENTATION: https://github.com/berkuva/TCGA-omics-integration.

SUMMARY: Identification of functional transcriptional regulators (TRs) associated with chromatin interactions is an important problem in studies of 3-dimensional genome organization and gene regulation. Direct inference of TR binding has been limited by the resolution of Hi-C data. Here, we present BART3D, a computational method for inferring TRs associated with genome-wide differential chromatin interactions by comparing Hi-C maps from two states, leveraging public ChIP-seq data for human and mouse. We demonstrate that BART3D can detect relevant TRs from dynamic Hi-C profiles with TR perturbation or cell differentiation. BART3D can be a useful tool in 3D genome data analysis and functional genomics research. AVAILABILITY AND IMPLEMENTATION: BART3D is implemented in Python and the source code is available at https://github.com/zanglab/bart3d. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Cancer has been characterized as a heterogeneous disease, and the classification of cancer subtypes has become a necessity in cancer research, as it can facilitate the subsequent clinical management of patients and provide clinical decision support for clinicians. With the advance of machine learning in the last decade, many researchers employ machine learning to tackle the cancer classification problem. Importantly, traditional machine learning algorithms require a large amount of annotated data for model training. However, collection of large amounts of annotated data is time-consuming and expensive and may not be realistic in real-world activities. Facing data scarcity, meta-learning is proposed to tackle this problem. Meta-learning utilizes prior knowledge learned from related tasks and generalizes to new tasks of limited supervised experience, and it has been applied in many fields to tackle scarce annotated data problem, such as few-shot image classification, drug discovery, etc. As data scarcity is common in cancer research and diagnosis studies, and there are only few previous studies that classify cancers based on limited annotated data. We explore the meta-learning algorithm (MAML) to tackle the scenario where only limited annotated data are available. In this work, our objective is to comprehensively compare MAML among few-shot learning methods (matching network and prototypical network) and traditional machine learning methods (random forest and K-nearest neighbor). Experimental results on The Cancer Genome Atlas (TCGA) cancer patient data demonstrates the effectiveness and superiority of MAML over other methods, including its ability to outperform the other methods using 4.5-fold fewer features.

MOTIVATION: Our previous study has shown that ERBB2 is overexpressed in the organoid model of MCF10A when the stiffness of the microenvironment is increased to that of high mammographic density (MD). We now aim to identify key transcription factors (TFs) and functional enhancers that regulate processes associated with increased stiffness of the microenvironment in the organoid models of premalignant human mammary cell lines. RESULTS: 3D colony organizations and the cis-regulatory networks of two human mammary epithelial cell lines (184A1 and MCF10A) are investigated as a function of the increased stiffness of the microenvironment within the range of MD. The 3D colonies are imaged using confocal microscopy, and the morphometries of colony organizations and heterogeneity are quantified as a function of the stiffness of the microenvironment using BioSig3D. In a surrogate assay, colony organizations are profiled by transcriptomics. Transcriptome data are enriched by correlative analysis with the computed morphometric indices. Next, a subset of enriched data are processed against publicly available ChIP-Seq data using Model-based Analysis of Regulation of Gene Expression to predict regulatory transcription factors. This integrative analysis of morphometric and transcriptomic data predicted YY1 as one of the cis-regulators in both cell lines as a result of the increased stiffness of the microenvironment. Subsequent experiments validated that YY1 is expressed at protein and mRNA levels for MCF10A and 184A1, respectively. Also, there is a causal relationship between activation of YY1 and ERBB2 when YY1 is overexpressed at the protein level in MCF10A. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Summary: Identification of functional transcription factors that regulate a given gene set is an important problem in gene regulation studies. Conventional approaches for identifying transcription factors, such as DNA sequence motif analysis, are unable to predict functional binding of specific factors and not sensitive enough to detect factors binding at distal enhancers. Here, we present binding analysis for regulation of transcription (BART), a novel computational method and software package for predicting functional transcription factors that regulate a query gene set or associate with a query genomic profile, based on more than 6000 existing ChIP-seq datasets for over 400 factors in human or mouse. This method demonstrates the advantage of utilizing publicly available data for functional genomics research. Availability and implementation: BART is implemented in Python and available at http://faculty.virginia.edu/zanglab/bart. Supplementary information: Supplementary data are available at Bioinformatics online.

BACKGROUND: Transcription factor binding, histone modification, and chromatin accessibility studies are important approaches to understanding the biology of gene regulation. ChIP-seq and DNase-seq have become the standard techniques for studying protein-DNA interactions and chromatin accessibility respectively, and comprehensive quality control (QC) and analysis tools are critical to extracting the most value from these assay types. Although many analysis and QC tools have been reported, few combine ChIP-seq and DNase-seq data analysis and quality control in a unified framework with a comprehensive and unbiased reference of data quality metrics. RESULTS: ChiLin is a computational pipeline that automates the quality control and data analyses of ChIP-seq and DNase-seq data. It is developed using a flexible and modular software framework that can be easily extended and modified. ChiLin is ideal for batch processing of many datasets and is well suited for large collaborative projects involving ChIP-seq and DNase-seq from different designs. ChiLin generates comprehensive quality control reports that include comparisons with historical data derived from over 23,677 public ChIP-seq and DNase-seq samples (11,265 datasets) from eight literature-based classified categories. To the best of our knowledge, this atlas represents the most comprehensive ChIP-seq and DNase-seq related quality metric resource currently available. These historical metrics provide useful heuristic quality references for experiment across all commonly used assay types. Using representative datasets, we demonstrate the versatility of the pipeline by applying it to different assay types of ChIP-seq data. The pipeline software is available open source at https://github.com/cfce/chilin . CONCLUSION: ChiLin is a scalable and powerful tool to process large batches of ChIP-seq and DNase-seq datasets. The analysis output and quality metrics have been structured into user-friendly directories and reports. We have successfully compiled 23,677 profiles into a comprehensive quality atlas with fine classification for users.

MOTIVATION: Chromatin states are the key to gene regulation and cell identity. Chromatin immunoprecipitation (ChIP) coupled with high-throughput sequencing (ChIP-Seq) is increasingly being used to map epigenetic states across genomes of diverse species. Chromatin modification profiles are frequently noisy and diffuse, spanning regions ranging from several nucleosomes to large domains of multiple genes. Much of the early work on the identification of ChIP-enriched regions for ChIP-Seq data has focused on identifying localized regions, such as transcription factor binding sites. Bioinformatic tools to identify diffuse domains of ChIP-enriched regions have been lacking. RESULTS: Based on the biological observation that histone modifications tend to cluster to form domains, we present a method that identifies spatial clusters of signals unlikely to appear by chance. This method pools together enrichment information from neighboring nucleosomes to increase sensitivity and specificity. By using genomic-scale analysis, as well as the examination of loci with validated epigenetic states, we demonstrate that this method outperforms existing methods in the identification of ChIP-enriched signals for histone modification profiles. We demonstrate the application of this unbiased method in important issues in ChIP-Seq data analysis, such as data normalization for quantitative comparison of levels of epigenetic modifications across cell types and growth conditions. AVAILABILITY: http://home.gwu.edu/ approximately wpeng/Software.htm. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.