Biology subjects

Greene, C. S.

Publications and source records attributed to Greene, C. S..

At least 19 recordsLinked to original sources

bioRxiv2026

Common cis-eQTLs do not explain chromosome 21 expression deviations from the ploidy expectation in Down syndrome

Down syndrome (DS) results from trisomy 21 (T21), the triplication of human chromosome 21 (HSA21). On average, genes on HSA21 are overexpressed at the expected 1.5-fold increase. A small subset of HSA21 genes is expressed closer to euploid levels. Could local genetic regulatory variation explain deviations from the expected dosage? Using the largest DS-specific transcriptomic dataset to date from the Human Trisome Project (HTP) and GTEx expression quantitative trait loci (eQTLs), we tested whether local genetic regulatory variation contributes to gene-specific buffering or amplification of HSA21 dosage effects in DS. We compared the bulk RNA-seq profiles of protein-coding genes, lncRNAs, and pseudogenes from whole-blood samples of 274 participants with T21 and 80 euploid (D21) participants. We filtered out genes with low read counts or high genomic repeat content and applied ploidy correction to the differential expression analysis. We found 13 genes that differed significantly from the expected fold change. We cross-referenced GTEx whole-blood eQTLs (cis window 1 {+/-} Mb) with the 13 deviating genes. A common cis-eQTL was detected for 8 of the 10 testable genes. However, 59 of 99 HSA21 genes that had the expected fold change had significant cis-eQTLs. Additionally, T21 participants carry the eQTL alleles at similar frequencies to GTEx donors. Therefore, cis-eQTLs explain differences between people with T21, not the difference between T21 and D21. Overall, common cis-regulatory variation acts on HSA21 genes in T21 as it does in euploid blood, but it does not explain which genes deviate from the ploidy expectation.

systems biology↗

bioRxiv2026

HetNetEX: Exact Asymptotic Inference in Heterogeneous Biomedical Knowledge Graphs

Heterogeneous biomedical knowledge networks (hetnets) integrate disparate data types, drugs, genes, diseases, and pathways, across independent sources; Hetionet (https://het.io) is a widely used example. A standard approach for assessing connectivity significance is XSwap, which permutes the hetnet P times and fits a gamma-hurdle null model to the degree-weighted path count (DWPC), pooling permuted values across pairs with matching source and target degrees to increase the effective sample size. This permutation approach has been highly successful in practice, but it faces four practical constraints in large graphs: (1) a finite resolution for the smallest reportable p-values, (2) computational cost that grows prohibitive at path lengths L [≥] 4 or 5, (3) a variance model (Var {propto} {micro}2) that departs from the configuration-model form (1 +{kappa} ){micro}, and (4) O(P 10m L) runtime. To complement this approach, we present HetNetEX (Heterogeneous Network EXact inference), which computes the null DWPC distribution analytically from degree sequences using the configuration model in O(Ln) time. In simulations at P = 200 across L = 1-4, HetNetEX achieves Spearman{rho} > 0.96 concordance with XSwap rankings while being >10,000x faster and providing analytical p-values without a resolution ceiling. High-degree pairs show larger XSwap sampling error than low-degree pairs, reflecting the finite-sample nature of permutation that analytical computation avoids.

systems biology↗

bioRxiv2026

Advances in protein function prediction from the fifth CAFA challenge

The Critical Assessment of Functional Annotation (CAFA) is a long-standing community effort to independently assess computational methods for protein function prediction, to highlight well-performing methodologies, to identify bottlenecks in the field, and to provide a forum for the dissemination of results and exchange of ideas. In its fifth round (CAFA5) of triennial challenges, a partnership with Kaggle Inc. facilitated participation from a large community of data scientists and computational biologists through a competitive prospective challenge on the crowdsourcing platform. In this work, we present an in-depth analysis of the submitted predictions and report improvements in accuracy over all methods from the previous CAFA challenges. We further introduce a new evaluation setting for proteins with pre-existing (incomplete) annotations and identify the need for methods that better leverage existing annotations to predict those that will be discovered later. Finally, we characterize the prospective evaluation framework by examining performance on a strict set of unpublished annotations and across intermediate database releases. Our results indicate that recent developments in the field, such as the availability of protein language models and accurately predicted 3D structures, as well as the growth of experimental annotations through biocuration, have all contributed to performance improvements. 1

bioinformatics↗

bioRxiv2026

Transcriptomic subtypes in high-grade serous ovarian cancer are driven by tumor cellular composition

High-grade serous ovarian carcinoma (HGSC) is an aggressive malignancy for which bulk transcriptomic subtypes are used to stratify tumors, interpret biology, and guide biomarker development. The four TCGA-derived subtypes, mesenchymal (C1.MES), immunoreactive (C2.IMM), proliferative (C5.PRO), and differentiated (C4.DIF), are consistently observed across cohorts. However, despite their prominence, these subtypes have not translated into therapeutic utility, and their biological basis remains unresolved. Here, we show that HGSC transcriptomic subtypes are largely determined by tumor cellular composition rather than intrinsic malignant transcriptional programs. By integrating controlled single-cell-derived pseudobulk simulations with deconvolution-based analysis of 1,834 primary HGSC tumors across RNA-seq and microarray cohorts, we demonstrate that subtype probabilities align along a composition-driven axis of stromal and immune variation. Cellular composition alone predicted subtype labels with high accuracy (ROC-AUC = 0.81-0.95) and explained a substantial fraction of subtype-associated transcriptomic variation, with the mesenchymal (C1.MES) subtype representing the most robust and reproducible example of composition-driven signal. Although a secondary, composition-independent expression signal is detectable, it does not define the dominant structure of subtype classification. These findings redefine HGSC transcriptomic subtypes as features of the tumor ecosystem rather than discrete malignant states. This reinterpretation has immediate implications for studies that use subtype labels to infer tumor-intrinsic biology and provides a generalizable framework for separating composition-driven and intrinsic signals in bulk tumor data. Significance StatementHGSC transcriptomic subtypes lack consistent clinical utility and remain biologically ambiguous. We show subtype assignments are largely driven by tumor cellular composition, and less so by distinct intrinsic tumor states.

cancer biology↗

bioRxiv2026

Deconvolved tumor adipocyte proportions and high grade serous ovarian carcinoma survival

BackgroundSingle-cell-based analyses of high-grade serous ovarian carcinoma (HGSOC) survival have largely ignored adipocytes, which are fragile and under-represented in single-cell references. Adipocytes are known active components of the tumor microenvironment in many cancers, and HGSOC tumors frequently metastasize to the omentum, a lining of adipose tissue. MethodsWe created a composite reference that combines single-nucleus adipose profiles with published HGSOC single-cell data to deconvolve 588 bulk RNA-seq tumours from the Schildkraut cohorts. We used stage-stratified Cox models to quantify the association between intratumoural adipocyte fractions and overall survival while adjusting for age, body mass index (BMI), race, and residual disease. We also evaluated associations with deconvolved immune, stromal, and epithelial cell groups. ResultsA 10% increase in estimated tumor adipocyte content was associated with a 41% increase in the hazard of death (HR = 1.41, 95% CI 1.18-1.70, p = 0.0002) after adjusting for age, BMI and race (n=566). A 10% increase in immune cell proportion was associated with favorable survival (HR = 0.82, 95% CI 0.69-0.97, p = 0.024). Stromal and epithelial macro-fractions were not associated with survival. Associations with adipocyte and immune cell type proportions were unchanged in models additionally controlling the other cell type proportions. Results were similar after additionally adjusting for residual disease after debulking surgery. ConclusionsAdipocytes may be a tumor-intrinsic factor associated with adverse outcomes in HGSOC. Quantifying adipocyte burden using bulk RNA-seq could enhance risk stratification and guide the development of adipocyte-targeted therapies.

genomics↗

bioRxiv2025

Characterizing intra- and inter-tumor heterogeneity in ovarian high-grade serous carcinoma subtypes using single-cell and spatial transcriptomics

Ovarian high-grade serous carcinoma (HGSC) is an aggressive ovarian cancer with a heterogeneous tumor microenvironment (TME). Advances in single-cell RNA sequencing (scRNA-seq) and spatially-resolved transcriptomics have enabled the study of complex TME. This study explores connections between molecular subtypes described from bulk transcriptomes and spatial domains characterized by distinct gene expression in HGSC and their variability between patients. We explored both intraand inter-tumor heterogeneity across 2D space and identified differing spatial patterns of gene expression pertaining to immune pathways and vasculature development. Functional characterization of tumor spaces revealed potentially shared cell states across molecular subtypes, while correlation analysis underscored subtype-specific spatial anti-colocalization between spots exhibiting antigen-presenting functions and B cell-mediated immunity. Lastly, we performed spatially-aware cell-cell communication analysis on the spatial samples and identified a molecular subtype specific difference in total signaling activity and heterogeneity in Midikine signaling between the differentiated subtype. Our results suggest that generating multiple tissue slices per patient might be necessary to enable comprehensive characterization of HGSC spatial transcriptomes.

genomics↗

bioRxiv2025

Accurate and robust classification of Mycobacterium bovis-infected cattle using peripheral blood RNA-seq data

Bovine tuberculosis (bTB) remains recalcitrant to eradication in many endemic countries where current diagnostics are suboptimal. Mycobacterium bovis causes bTB and is closely related to Mycobacterium tuberculosis, which causes human tuberculosis (hTB). Although blood-based mRNA biomarkers identified through machine learning can discriminate hTB-positive from hTB-negative individuals, similar approaches have not been explored for bTB. Here, we use RNA-seq and machine learning to investigate the utility of blood mRNA as a host-response biomarker for bTB. We identify a 30-gene signature and a 273-gene elastic net classifier that differentiate bTB-positive from bTB-negative cattle, achieving area under the curve (AUC) values of 0.986/0.900 for the former and 0.968/0.938 for the latter in training and testing, respectively. Additionally, we show that these classifiers distinguish bTB-positive cattle from cattle infected with other microbial pathogens (AUC [≥] 0.819). These mRNA-based classifiers represent a promising tool for augmenting current diagnostics to advance global bTB eradication efforts.

genomics↗

bioRxiv2025

Integrating single-cell and single-nucleus datasets improves bulk RNA-seq deconvolution

Bulk RNA-seq deconvolution typically uses single-cell RNA-sequencing (scRNA-seq) references, but some cell types are only detectable through single-nucleus RNA sequencing (snRNA-seq). Because snRNA-seq captures nuclear, but not cytoplasmic, transcripts, direct use as a reference could reduce deconvolution accuracy. Here, we systematically benchmark strategies to integrate both modalities, focusing on transformations and gene-filtering approaches that harmonize snRNA-seq with scRNA-seq references. Across four diverse tissues, we evaluated principal component-based shifts, conditional and non-conditional variational autoencoders (scVI), and the removal of cross-modality differentially expressed genes (DEGs). While all methods improved performance relative to untransformed snRNA-seq, filtering consistent cross-modality DEGs delivered the greatest gains, often matching or surpassing scRNA-only references. Conditional scVI performed comparably and was especially effective when matched scRNA-snRNA cell types were unavailable. In real adipose bulk samples without ground truth, DEG pruning and conditional scVI provided the most robust cell-fraction estimates across donors and transformations. Together, these results demonstrate that scRNA-seq should be prioritized as the reference when available, with snRNA-seq appended only after filtering cross-modality DEGs. For less-characterized systems where DEG information is limited, conditional scVI offers a practical alternative. Our findings provide clear guidelines for modality-aware integration, enabling near-scRNA-seq accuracy in bulk deconvolution workflows.

bioinformatics↗

bioRxiv2024

Missing cell types in single-cell references impact deconvolution of bulk data but are detectable

Advancements in RNA-sequencing have dramatically expanded our ability to study gene expression profiles of biological samples in bulk tissue and single cells. Deconvolution of bulk data with single-cell references provides the ability to study relative cell-type proportions, but most methods assume a reference is present for every cell type in bulk data. This is not true in all circumstances--cell types can be missing in single-cell profiles for many reasons. In this study, we examine the impact of missing cell types on deconvolution methods. Our experimental designs are simulation-based, using paired single-cell and single-nucleus data, since single-nucleus RNA-sequencing is able to preserve the nucleus of cell types that would otherwise be missing in a single-cell counterpart. These datasets allow us to examine the missing-cell-type phenomenon in deconvolution with realistic proportions. We apply three deconvolution methods that vary from straightforward to state-of-the-art: non-negative least squares, BayesPrism, and CIBERSORTx. We find that the performance of deconvolution methods is influenced by both the number and the similarity of missing cell types, consistent with prior results. Additionally, we find that missing cell-type profiles can be recovered from residuals using a simple non-negative matrix factorization strategy. We expect our simulation strategies and results to provide a starting point for those developing new deconvolution methods and help improve their to better account for the presence of missing cell types. Building off of our findings on simulated data, we then analyzed data from high-grade serous ovarian cancer; a tumor that has regions of highly variable levels of adipocytes dependent on the region from which it is sampled. We observe results consistent with simulation, namely that expression patterns from cell types likely to be missing appear present in residuals. Our results suggests that deconvolution methods should consider the possibility of missing cell types and provide a starting point to address this. Our source code for data simulation and analysis is freely available at https://github.com/greenelab/pred_missing_celltypes.

bioinformatics↗

bioRxiv2024

The Single-cell Pediatric Cancer Atlas: Data portal and open-source tools for single-cell transcriptomics of pediatric tumors

The Single-cell Pediatric Cancer Atlas (ScPCA) Portal (https://scpca.alexslemonade.org/) is a data resource for uniformly processed single-cell and single-nuclei RNA sequencing (RNA-seq) data and de-identified metadata from pediatric tumor samples. Originally comprised of data from 10 projects funded by Alexs Lemonade Stand Foundation (ALSF), the Portal currently contains summarized gene expression data for over 700 samples across 55 cancer types from ALSF-funded and community-contributed datasets. Downloads include gene expression data as SinglecellExperiment or AnnData objects containing raw and normalized counts, PCA and UMAP coordinates, and automated cell type annotations, along with summary reports. Some samples have additional data from bulk RNA-seq, spatial transcriptomics, and/or feature barcoding (e.g., CITE-seq and cell hashing) included in the download. All data on the Portal were uniformly processed using scpca-nf, an efficient and open-source Nextflow workflow that uses alevin-fry to quantify gene expression. Comprehensive documentation, including descriptions of file contents and a guide to getting started, is available at https://scpca.readthedocs.io.

bioinformatics↗

bioRxiv2023

Molecular subtypes of high grade serous ovarian cancer across racial groups and gene expression platforms

IntroductionHigh-grade serous carcinoma (HGSC) gene expression subtypes are associated with differential survival. We characterized HGSC gene expression in Black individuals and considered whether gene expression differences by race may contribute to poorer HGSC survival among Black versus non-Hispanic White individuals. MethodsWe included newly generated RNA-Seq data from Black and White individuals, and array-based genotyping data from four existing studies of White and Japanese individuals. We assigned subtypes using K-means clustering. Cluster- and dataset-specific gene expression patterns were summarized by moderated t-scores. We compared cluster-specific gene expression patterns across datasets by calculating the correlation between the summarized vectors of moderated t-scores. Following mapping to The Cancer Genome Atlas (TCGA)-derived HGSC subtypes, we used Cox proportional hazards models to estimate subtype-specific survival by dataset. ResultsCluster-specific gene expression was similar across gene expression platforms. Comparing the Black study population to the White and Japanese study populations, the immunoreactive subtype was more common (39% versus 23%-28%) and the differentiated subtype less common (7% versus 22%-31%). Patterns of subtype-specific survival were similar between the Black and White populations with RNA-Seq data; compared to mesenchymal cases, the risk of death was similar for proliferative and differentiated cases and suggestively lower for immunoreactive cases (Black population HR=0.79 [0.55, 1.13], White population HR=0.86 [0.62, 1.19]). ConclusionsA single, platform-agnostic pipeline can be used to assign HGSC gene expression subtypes. While the observed prevalence of HGSC subtypes varied by race, subtype-specific survival was similar. Statement of SignificanceA single pipeline was used to subtype ovarian high-grade serous carcinoma (HGSC) with array-based or RNA-Seq gene expression data. Subtype distributions differed by race, but subtype-specific survival was similar across racial groups.

cancer biology↗

bioRxiv2023

Integration of 168,000 samples reveals global patterns of the human gut microbiome

Understanding the factors that shape variation in the human microbiome is a major goal of research in biology. While other genomics fields have used large, pre-compiled compendia to extract systematic insights requiring otherwise impractical sample sizes, there has been no comparable resource for the 16S rRNA sequencing data commonly used to quantify microbiome composition. To help close this gap, we have assembled a set of 168,484 publicly available human gut microbiome samples, processed with a single pipeline and combined into the largest unified microbiome dataset to date. We use this resource, which is freely available at microbiomap.org, to shed light on global variation in the human gut microbiome. We find that Firmicutes, particularly Bacilli and Clostridia, are almost universally present in the human gut. At the same time, the relative abundance of the 65 most common microbial genera differ between at least two world regions. We also show that gut microbiomes in undersampled world regions, such as Central and Southern Asia, differ significantly from the more thoroughly characterized microbiomes of Europe and Northern America. Moreover, humans in these overlooked regions likely harbor hundreds of taxa that have not yet been discovered due to this undersampling, highlighting the need for diversity in microbiome studies. We anticipate that this new compendium can serve the community and enable advanced applied and methodological research.

microbiology↗

bioRxiv2023

Analysis of Pseudomonas aeruginosa transcription in an ex vivo cystic fibrosis sputum model identifies metal restriction as a gene expression stimulus

Chronic Pseudomonas aeruginosa lung infections are a distinctive feature of cystic fibrosis (CF) pathology, that challenge adults with CF even with the advent of highly effective modulator therapies. Characterizing P. aeruginosa transcription in the CF lung and identifying factors that drive gene expression could yield novel strategies to eradicate infection or otherwise improve outcomes. To complement published P. aeruginosa gene expression studies in laboratory culture models designed to model the CF lung environment, we employed an ex vivo sputum model in which laboratory strain PAO1 was incubated in sputum from different CF donors. As part of the analysis, we compared PAO1 gene expression in this "spike-in" sputum model to that for P. aeruginosa grown in artificial sputum medium (ASM). Analyses focused on genes that were differentially expressed between sputum and ASM and genes that were most highly expressed in sputum. We present a new approach that used sets of genes with correlated expression, identified by the gene expression analysis tool eADAGE, to analyze the differential activity of pathways in P. aeruginosa grown in CF sputum from different individuals. A key characteristic of P. aeruginosa grown in expectorated CF sputum was related to zinc and iron acquisition, but this signal varied by donor sputum. In addition, a significant correlation between P. aeruginosa expression of the H1-type VI secretion system and corrector use by the sputum donor was observed. These methods may be broadly useful in looking for variable signals across clinical samples. ImportanceIdentifying the gene expression programs used by Pseudomonas aeruginosa to colonize the lungs of people with cystic fibrosis (CF) will illuminate new therapeutic strategies. To capture these transcriptional programs, we cultured the common P. aeruginosa laboratory strain PAO1 in expectorated sputum from CF patient donors. Through bioinformatics analysis, we defined sets of genes that are more transcriptionally active in real CF sputum compared to artificial sputum media (ASM). Many of the most differentially active gene sets contained genes related to metal acquisition, suggesting that these gene sets play an active role in scavenging for metals in the CF lung environment which is inadequately represented in ASM. Future studies of P. aeruginosa transcription in CF may benefit from the use of an expectorated sputum model or modified forms of ASM supplemented with metals.

microbiology↗

bioRxiv2023

MousiPLIER: A Mouse Pathway-Level Information Extractor Model

High throughput gene expression profiling is a powerful approach to generate hypotheses on the underlying causes of biological function and disease. Yet this approach is limited by its ability to infer underlying biological pathways and burden of testing tens of thousands of individual genes. Machine learning models that incorporate prior biological knowledge are necessary to extract meaningful pathways and generate rational hypothesis from the vast amount of gene expression data generated to date. We adopted an unsupervised machine learning method, Pathway-level information extractor (PLIER), to train the first mouse PLIER model on 190,111 mouse brain RNA-sequencing samples, the greatest amount of training data ever used by PLIER. mousiPLER converted gene expression data into a latent variables that align to known pathway or cell maker gene sets, substantially reducing data dimensionality and improving interpretability. To determine the utility of mousiPLIER, we applied it to a mouse brain aging study of microglia and astrocyte transcriptomic profiling. We found a specific set of latent variables that are significantly associated with aging, including one latent variable (LV41) corresponding to striatal signal. We next performed k-means clustering on the training data to identify studies that respond strongly to LV41, finding that the variable is relevant to striatum and aging across the scientific literature. Finally, we built a web server (http://mousiplier.greenelab.com/) for users to easily explore the learned latent variables. Taken together this study provides proof of concept that mousiPLIER can uncover meaningful biological processes in mouse transcriptomic studies. Significance statementAnalysis of RNA-sequencing data commonly generates differential expression of individual genes across conditions. However, genes are regulated in complex networks, not as individual entities. Machine learning models that incorporate prior biological information are a powerful tool to analyze human gene expression. However, such models are lacking for mouse despite the vast number of mouse RNA-seq datasets. We trained a mouse pathway-level information extractor model (mousiPLIER). The model reduced the data dimensionality from over 10,000 genes to 196 latent variables that map to prior pathway and cell marker gene sets. We demonstrated the utility of mousiPLIER by applying it to mouse brain aging data and developed a web server to facilitate the use of the model by the scientific community.

bioinformatics↗

bioRxiv2023

BuDDI: Bulk Deconvolution with Domain Invariance to predict cell-type-specific perturbations from bulk

While single-cell experiments provide deep cellular resolution within a single sample, some single-cell experiments are inherently more challenging than bulk experiments due to dissociation difficulties, cost, or limited tissue availability. This creates a situation where we have deep cellular profiles of one sample or condition, and bulk profiles across multiple samples and conditions. To bridge this gap, we propose BuDDI (BUlk Deconvolution with Domain Invariance). BuDDI utilizes domain adaptation techniques to effectively integrate available corpora of case-control bulk and reference scRNA-seq observations to infer cell-type-specific perturbation effects. BuDDI achieves this by learning independent latent spaces within a single variational autoencoder (VAE) encompassing at least four sources of variability: 1) cell type proportion, 2) perturbation effect, 3) structured experimental variability, and 4) remaining variability. Since each latent space is encouraged to be independent, we simulate perturbation responses by independently composing each latent space to simulate cell-type-specific perturbation responses. We evaluated BuDDIs performance on simulated and real data with experimental designs of increasing complexity. We first validated that BuDDI could learn domain invariant latent spaces on data with matched samples across each source of variability. Then we validated that BuDDI could accurately predict cell-type-specific perturbation response when no single-cell perturbed profiles were used during training; instead, only bulk samples had both perturbed and non-perturbed observations. Finally, we validated BuDDI on predicting sex-specific differences, an experimental design where it is not possible to have matched samples. In each experiment, BuDDI outperformed all other comparative methods and baselines. As more reference atlases are completed, BuDDI provides a path to combine these resources with bulk-profiled treatment or disease signatures to study perturbations, sex differences, or other factors at single-cell resolution.

bioinformatics↗

bioRxiv2023

Is a Picture Worth 1,000 SNPs? Effects of User-Submitted Photographs on Ancestry Estimates from Direct-to-Consumer Canine Genetic Tests

ObjectiveTo evaluate whether the breed ancestry predictions of direct-to-consumer (DTC) genetic tests for dogs are influenced by the user-provided photograph. AnimalsTwelve pet dogs considered purebred (i.e., registered with a breed organization) representing twelve different breeds. MethodsSix buccal swabs per dog were collected by the owners and submitted to six DTC genetic testing companies. The experimenters registered each sample with the company. For half of the dogs, the registration included a photograph of the DNA donor. For the other half of the dogs, photographs were swapped between dogs. Analysis of the DNA and breed ancestry prediction was conducted by each company. Each companys breed predictions were evaluated to assess whether the condition (i.e., matching versus shuffled photograph) affected the odds of identifying the DNA donors registered breed. A convolutional neural network was also used to predict breed based solely on the photograph as a positive control. ResultsFive of the six tests always produced results that included the registered breed. One test and the convolutional neural network were unlikely to identify the registered breed and frequently returned results that included the breed in the photograph. This result suggests that one test on the market is relying on the photograph more than the DNA sample. Additionally, differences in the predictions made across all tests underscore the challenge of identifying breed ancestry, even in purebred dogs. Clinical RelevanceVeterinarians are likely to encounter patients who have conducted DTC genetic testing and may find themselves in the position of explaining genetic test results that they did not order. This systematic comparison of tests on the market provides context for interpreting unexpected results from consumer-grade DTC genetic testing kits.

genetics↗

bioRxiv2023

Optimizer's dilemma: optimization strongly influences model selection in transcriptomic prediction

MotivationMost models can be fit to data using various optimization approaches. While model choice is frequently reported in machine-learning-based research, optimizers are not often noted. We applied two different implementations of LASSO logistic regression implemented in Pythons scikit-learn package, using two different optimization approaches (coordinate descent and stochastic gradient descent), to predict driver mutation presence or absence from gene expression across 84 pan-cancer driver genes. Across varying levels of regularization, we compared performance and model sparsity between optimizers. ResultsAfter model selection and tuning, we found that coordinate descent (implemented in the liblinear library) and SGD tended to perform comparably. liblinear models required more extensive tuning of regularization strength, performing best for high model sparsities (more nonzero coefficients), but did not require selection of a learning rate parameter. SGD models required tuning of the learning rate to perform well, but generally performed more robustly across different model sparsities as regularization strength decreased. Given these tradeoffs, we believe that the choice of optimizers should be clearly reported as a part of the model selection and validation process, to allow readers and reviewers to better understand the context in which results have been generated. Availability and implementationThe code used to carry out the analyses in this study is available at https://github.com/greenelab/pancancer-evaluation/tree/master/01_stratified_classification. Performance/regularization strength curves for all genes in the Vogelstein et al. 2013 dataset are available at https://doi.org/10.6084/m9.figshare.22728644.

bioinformatics↗

bioRxiv2023

Deconvolution reveals compositional differences in high-grade serous ovarian cancer subtypes

Ovarian cancer is a deadly disease with few effective therapies. The most common form is high-grade serous ovarian cancer (HGSOC). Transcriptomic subtypes of HGSOC have shown promise in characterizing tumor heterogeneity and are associated with survival. Gene expression signatures for the subtypes suggest variation in stromal cell types in the tumor microenvironment (TME). Here, we characterize the TME composition of HGSOC on a population scale by performing deconvolution on bulk transcriptomic data. We use comprehensive cell type profiles from 164 HGSOC tumor samples from two independent reference datasets, in order to compare cell type proportions across and within bulk transcriptomic datasets, and assess their alignment to the subtypes proposed by The Cancer Genome Atlas. We also assess the relationship between tumor composition and clinical outcomes. Our results suggest that HGSOC transcriptomic subtypes are driven by TME composition, specifically fibroblast and immune cell content, and we propose a modified HGSOC subtype model informed by cell composition.

bioinformatics↗