The integration of single-cell multi-omics data provides a powerful approach for understanding the complex interplay between different molecular modalities, such as RNA expression, chromatin accessibility and protein abundance, measured through assays like scRNA-seq, scATAC-seq and CITE-seq, at single-cell resolution. However, most existing single-cell technologies focus on individual modalities, limiting a comprehensive understanding of their interconnections. Integrating such diverse and often unpaired datasets remains a challenging task due to unknown cell correspondences across distinct feature spaces and limited insights into cell-type-specific activities in non-scRNA-seq modalities. In this work, we propose BiCLUM, a Bilateral Contrastive Learning approach for Unpaired single-cell Multi-omics integration, which simultaneously enforces cell-level and feature-level alignment across modalities. BiCLUM first transforms one modality, such as scATAC-seq, into the data space of another modality, such as scRNA-seq, using prior genomic knowledge. It then learns cell and gene embeddings simultaneously through a bilateral contrastive learning framework, incorporating both cell-level and feature-level contrastive losses. Across multiple RNA+ATAC and RNA+protein datasets, BiCLUM consistently outperforms or matches existing integration methods in both visualization and quantitative benchmarks. Importantly, BiCLUM embeddings preserve biologically meaningful regulatory relationships between chromatin accessibility and gene expression, as evidenced by significantly higher gene-peak correlations than random controls. Downstream analyses further demonstrate that BiCLUM-derived embeddings facilitate transcription factor activity inference, identification of cell-type-specific marker genes, functional enrichment, and cell-cell interaction mapping. Comprehensive hyperparameter sensitivity and ablation analyses further establish BiCLUM as a robust and interpretable framework that not only achieves effective cross-modal alignment but also retains the underlying regulatory and functional landscape across single-cell modalities.

Benchmarking, which involves collecting reference datasets and demonstrating method performance, is a requirement for the development of new computational tools, but also becomes a domain of its own to achieve neutral comparisons of methods. Although a lot has been written about how to design and conduct benchmark studies, this Perspective sheds light on a wish list for a computational platform to orchestrate benchmark studies. We discuss various ideas for organizing reproducible software environments, formally defining benchmarks, orchestrating standardized workflows, and how they interface with computing infrastructure.

MOTIVATION: Spatially resolved transcriptomics (SRT) enables scientists to investigate spatial context of mRNA abundance, including identifying spatially variable genes (SVGs), i.e. genes whose expression varies across the tissue. Although several methods have been proposed for this task, native SVG tools cannot jointly model biological replicates, or identify the key areas of the tissue affected by spatial variability. RESULTS: Here, we introduce DESpace, a framework, based on an original application of existing methods, to discover SVGs. In particular, our approach inputs all types of SRT data, summarizes spatial information via spatial clusters, and identifies spatially variable genes by performing differential gene expression testing between clusters. Furthermore, our framework can identify (and test) the main cluster of the tissue affected by spatial variability; this allows scientists to investigate spatial expression changes in specific areas of interest. Additionally, DESpace enables joint modeling of multiple samples (i.e. biological replicates); compared to inference based on individual samples, this approach increases statistical power, and targets SVGs with consistent spatial patterns across replicates. Overall, in our benchmarks, DESpace displays good true positive rates, controls for false positive and false discovery rates, and is computationally efficient. AVAILABILITY AND IMPLEMENTATION: DESpace is freely distributed as a Bioconductor R package at https://bioconductor.org/packages/DESpace.

ATAC-seq has emerged as a rich epigenome profiling technique, and is commonly used to identify Transcription Factors (TFs) underlying given phenomena. A number of methods can be used to identify differentially-active TFs through the accessibility of their DNA-binding motif, however little is known on the best approaches for doing so. Here we benchmark several such methods using a combination of curated datasets with various forms of short-term perturbations on known TFs, as well as semi-simulations. We include both methods specifically designed for this type of data as well as some that can be repurposed for it. We also investigate variations to these methods, and identify three particularly promising approaches (a chromVAR-limma workflow with critical adjustments, monaLisa and a combination of GC smooth quantile normalization and multivariate modeling). We further investigate the specific use of nucleosome-free fragments, the combination of top methods, and the impact of technical variation. Finally, we illustrate the use of the top methods on a novel dataset to characterize the impact on DNA accessibility of TRAnscription Factor TArgeting Chimeras (TRAFTAC), which can deplete TFs-in our case NFkB-at the protein level.

Computational biologists are frequently engaged in collaborative data analysis with wet lab researchers. These interdisciplinary projects, as necessary as they are to the scientific endeavor, can be surprisingly challenging due to cultural differences in operations and values. In this Ten Simple Rules guide, we aim to help dry lab researchers identify sources of friction and provide actionable tools to facilitate respectful, open, transparent, and rewarding collaborations.

BACKGROUND: Innovations in single cell technologies have lead to a flurry of datasets and computational tools to process and interpret them, including analyses of cell composition changes and transition in cell states. The diffcyt workflow for differential discovery in cytometry data consist of several steps, including preprocessing, cell population identification and differential testing for an association with a binary or continuous covariate. However, the commonly measured quantity of survival time in clinical studies often results in a censored covariate where classical differential testing is inapplicable. RESULTS: To overcome this limitation, multiple methods to directly include censored covariates in differential abundance analysis were examined with the use of simulation studies and a case study. Results show that multiple imputation based methods offer on-par performance with the Cox proportional hazards model in terms of sensitivity and error control, while offering flexibility to account for covariates. The tested methods are implemented in the R package censcyt as an extension of diffcyt and are available at https://bioconductor.org/packages/censcyt . CONCLUSION: Methods for the direct inclusion of a censored variable as a predictor in GLMMs are a valid alternative to classical survival analysis methods, such as the Cox proportional hazard model, while allowing for more flexibility in the differential analysis.

Summary: Statistical tools for biological data analysis are often evaluated using synthetic data, designed to mimic the features of a specific type of experimental data. The generalizability of such evaluations depends on how well the synthetic data reproduce the main characteristics of the experimental data, and we argue that an assessment of this similarity should accompany any synthetic dataset used for method evaluation. We describe countsimQC, which provides a straightforward way to generate a stand-alone report that shows the main characteristics of (e.g. RNA-seq) count data and can be provided alongside a publication as verification of the appropriateness of any utilized synthetic data. Availability and implementation: countsimQC is implemented as an R package (for R versions ≥ 3.4) and is available from https://github.com/csoneson/countsimQC under a GPL (≥2) license. Contact: [email protected] or [email protected].

SUMMARY: It is expected that emerging digital gene expression (DGE) technologies will overtake microarray technologies in the near future for many functional genomics applications. One of the fundamental data analysis tasks, especially for gene expression studies, involves determining whether there is evidence that counts for a transcript or exon are significantly different across experimental conditions. edgeR is a Bioconductor software package for examining differential expression of replicated count data. An overdispersed Poisson model is used to account for both biological and technical variability. Empirical Bayes methods are used to moderate the degree of overdispersion across transcripts, improving the reliability of inference. The methodology can be used even with the most minimal levels of replication, provided at least one phenotype or experimental condition is replicated. The software may have other applications beyond sequencing data, such as proteome peptide count data. AVAILABILITY: The package is freely available under the LGPL licence from the Bioconductor web site (http://bioconductor.org).

SUMMARY: Epigenetics, the study of heritable somatic phenotypic changes not related to DNA sequence, has emerged as a critical component of the landscape of gene regulation. The epigenetic layers, such as DNA methylation, histone modifications and nuclear architecture are now being extensively studied in many cell types and disease settings. Few software tools exist to summarize and interpret these datasets. We have created a toolbox of procedures to interrogate and visualize epigenomic data (both array- and sequencing-based) and make available a software package for the cross-platform R language. AVAILABILITY: The package is freely available under LGPL from the R-Forge web site (http://repitools.r-forge.r-project.org/) CONTACT: [email protected].

BACKGROUND: The latest generation of Affymetrix microarrays are designed to interrogate expression over the entire length of every locus, thus giving the opportunity to study alternative splicing genome-wide. The Exon 1.0 ST (sense target) platform, with versions for Human, Mouse and Rat, is designed primarily to probe every known or predicted exon. The smaller Gene 1.0 ST array is designed as an expression microarray but still interrogates expression with probes along the full length of each well-characterized transcript. We explore the possibility of using the Gene 1.0 ST platform to identify differential splicing events. RESULTS: We propose a strategy to score differential splicing by using the auxiliary information from fitting the statistical model, RMA (robust multichip analysis). RMA partitions the probe-level data into probe effects and expression levels, operating robustly so that if a small number of probes behave differently than the rest, they are downweighted in the fitting step. We argue that adjacent poorly fitting probes for a given sample can be evidence of differential splicing and have designed a statistic to search for this behaviour. Using a public tissue panel dataset, we show many examples of tissue-specific alternative splicing. Furthermore, we show that evidence for putative alternative splicing has a strong correspondence between the Gene 1.0 ST and Exon 1.0 ST platforms. CONCLUSION: We propose a new approach, FIRMAGene, to search for differentially spliced genes using the Gene 1.0 ST platform. Such an analysis complements the search for differential expression. We validate the method by illustrating several known examples and we note some of the challenges in interpreting the probe-level data.Software implementing our methods is freely available as an R package.

MOTIVATION: Analyses of EST data show that alternative splicing is much more widespread than once thought. The advent of exon and tiling microarrays means that researchers now have the capacity to experimentally measure alternative splicing on a genome wide level. New methods are needed to analyze the data from these arrays. RESULTS: We present a method, finding isoforms using robust multichip analysis (FIRMA), for detecting differential alternative splicing in exon array data. FIRMA has been developed for Affymetrix exon arrays, but could in principle be extended to other exon arrays, tiling arrays or splice junction arrays. We have evaluated the method using simulated data, and have also applied it to two datasets: a panel of 11 human tissues and a set of 10 pairs of matched normal and tumor colon tissue. FIRMA is able to detect exons in several genes confirmed by reverse transcriptase PCR. AVAILABILITY: R code implementing our methods is contributed to the package aroma.affymetrix.

MOTIVATION: Digital gene expression (DGE) technologies measure gene expression by counting sequence tags. They are sensitive technologies for measuring gene expression on a genomic scale, without the need for prior knowledge of the genome sequence. As the cost of sequencing DNA decreases, the number of DGE datasets is expected to grow dramatically. Various tests of differential expression have been proposed for replicated DGE data using binomial, Poisson, negative binomial or pseudo-likelihood (PL) models for the counts, but none of the these are usable when the number of replicates is very small. RESULTS: We develop tests using the negative binomial distribution to model overdispersion relative to the Poisson, and use conditional weighted likelihood to moderate the level of overdispersion across genes. Not only is our strategy applicable even with the smallest number of libraries, but it also proves to be more powerful than previous strategies when more libraries are available. The methodology is equally applicable to other counting technologies, such as proteomic spectral counts. AVAILABILITY: An R package can be accessed from http://bioinf.wehi.edu.au/resources/

BACKGROUND: Affymetrix GeneChips are an important tool in many facets of biological research. Recently, notable design changes to the chips have been made. In this study, we use publicly available data from Affymetrix to gauge the performance of three human gene expression arrays: Human Genome U133 Plus 2.0 (U133), Human Exon 1.0 ST (HuEx) and Human Gene 1.0 ST (HuGene). RESULTS: We studied probe-, exon- and gene-level reproducibility of technical and biological replicates from each of the 3 platforms. The U133 array has larger feature sizes so it is no surprise that probe-level variances are smaller, however the larger number of probes per gene on the HuGene array seems to produce gene-level summaries that have similar variances. The gene-level summaries of the HuEx array are less reproducible than the other two, despite having the largest average number of probes per gene. Greater than 80% of the content on the HuEx arrays is expressed at or near background. Biological variation seems to have a smaller effect on U133 data. Comparing the overlap of differentially expressed genes, we see a high overall concordance among all 3 platforms, with HuEx and HuGene having greater overlap, as expected given their design. We performed an analysis of detection rates and area under ROC curves using an experiment made up of several mixtures of 2 human tissues. Though it appears that the HuEx array has worse performance in terms of detection rates, all arrays have similar ability to separate differentially expressed and non-differentially expressed genes. CONCLUSION: Despite noticeable differences in the probe-level reproducibility, gene-level reproducibility and differential expression detection are quite similar across the three platforms. The HuEx array, an all-encompassing array, has the flexibility of measuring all known or predicted exonic content. However, the HuEx array induces poorer reproducibility for genes with fewer exons. The HuGene measures just the well-annotated genome content and appears to perform well. The U133 array, though not able to measure across the full length of a transcript, appears to perform as well as the newer designs on the set of genes common to all 3 platforms.

BACKGROUND: Gas chromatography-mass spectrometry (GC-MS) is a robust platform for the profiling of certain classes of small molecules in biological samples. When multiple samples are profiled, including replicates of the same sample and/or different sample states, one needs to account for retention time drifts between experiments. This can be achieved either by the alignment of chromatographic profiles prior to peak detection, or by matching signal peaks after they have been extracted from chromatogram data matrices. Automated retention time correction is particularly important in non-targeted profiling studies. RESULTS: A new approach for matching signal peaks based on dynamic programming is presented. The proposed approach relies on both peak retention times and mass spectra. The alignment of more than two peak lists involves three steps: (1) all possible pairs of peak lists are aligned, and similarity of each pair of peak lists is estimated; (2) the guide tree is built based on the similarity between the peak lists; (3) peak lists are progressively aligned starting with the two most similar peak lists, following the guide tree until all peak lists are exhausted. When two or more experiments are performed on different sample states and each consisting of multiple replicates, peak lists within each set of replicate experiments are aligned first (within-state alignment), and subsequently the resulting alignments are aligned themselves (between-state alignment). When more than two sets of replicate experiments are present, the between-state alignment also employs the guide tree. We demonstrate the usefulness of this approach on GC-MS metabolic profiling experiments acquired on wild-type and mutant Leishmania mexicana parasites. CONCLUSION: We propose a progressive method to match signal peaks across multiple GC-MS experiments based on dynamic programming. A sensitive peak similarity function is proposed to balance peak retention time and peak mass spectra similarities. This approach can produce the optimal alignment between an arbitrary number of peak lists, and models explicitly within-state and between-state peak alignment. The accuracy of the proposed method was close to the accuracy of manually-curated peak matching, which required tens of man-hours for the analyzed data sets. The proposed approach may offer significant advantages for processing of high-throughput metabolomics data, especially when large numbers of experimental replicates and multiple sample states are analyzed.

BACKGROUND: For effective exposition of biological information, especially with regard to analysis of large-scale data types, researchers need immediate access to multiple categorical knowledge bases and need summary information presented to them on collections of genes, as opposed to the typical one gene at a time. RESULTS: We present here a web-based tool (FunSpec) for statistical evaluation of groups of genes and proteins (e.g. co-regulated genes, protein complexes, genetic interactors) with respect to existing annotations (e.g. functional roles, biochemical properties, localization). FunSpec is available online at http://funspec.med.utoronto.ca CONCLUSION: FunSpec is helpful for interpretation of any data type that generates groups of related genes and proteins, such as gene expression clustering and protein complexes, and is useful for predictive methods employing "guilt-by-association."