Casey S. Greene

wenda_gpu provides a fast domain adaptation method for genomic data using GPU acceleration, achieving comparable results to the original wenda while improving cancer mutation prediction on small datasets.

This study compares neural networks and logistic regression for transcriptomic prediction, finding both linear and non-linear signals are present but not always sufficient for neural networks to outperform linear models.

The paper analyzes transcriptional patterns of core and accessory genes in P. aeruginosa strains using a transcriptome compendium, revealing stable core gene expression patterns across strains.

The paper evaluates the effectiveness of various -omics data types in predicting cancer mutation states and finds that RNA sequencing is often the best, but other data types can be equally effective.

The paper uses multi-ancestry biobank data and integrative methods to identify gene-disease associations across different ancestry groups, finding shared and ancestry-specific genetic links.

The paper evaluates genetic demultiplexing tools for pooled single-cell RNA-seq in cancer, showing they work effectively despite challenges from ambient RNA.

miQC is a probabilistic framework that jointly models mitochondrial DNA and gene detection metrics to improve single-cell RNA-seq quality control.

GenomicSuperSignature compares new RNA-seq datasets to public archives for rapid interpretation using precomputed axes of variation.

This paper introduces SOPHIE, a method using generative neural networks to distinguish common from specific transcriptional responses across datasets.

The paper analyzes linguistic features of bioRxiv preprints compared to published biomedical texts and develops tools to link preprints to peer-reviewed articles and journals.

The paper identifies 40 novel genetic loci associated with longitudinal pediatric bone accrual and identifies candidate effector genes through variant-to-gene mapping and functional validation.

The paper simulates gene expression compendia to show that technical variability masks biological signals, but correction methods help when few experiments are combined, while more experiments make detection easier without correction.

The paper analyzes disparities in recognition of computational biology researchers based on gender, name origin, country, and race/ethnicity.

The paper identifies histone modifications associated with alternative exon selection during mammalian development using machine learning.

The paper uses statistical and machine learning methods to analyze NF1 tumor genomic data, identifying key signaling pathways and molecular features associated with different tumor types and potential drug targets.

The paper shows that removing low-degree nodes from a gene interaction network improves prediction of expression dependencies compared to a random model.

This paper genomically characterizes 261 pediatric cancer patient-derived xenograft models to guide rational clinical trial design and drug development.

The paper evaluates multiple compression methods and dimensionalities on gene expression data, finding no single best approach but showing benefits from ensemble methods across different dimensions.

The paper generates synthetic clinical data using deep neural networks with differential privacy to enable privacy-preserving data sharing for secondary analyses.

MultiPLIER uses transfer learning from large public data to improve analysis of small rare disease datasets by capturing systemic biological features.

The paper identifies 11 novel Drosophila genes involved in long-term memory formation using computational methods and behavioral assays.

The paper evaluates the Tybalt VAE for scRNA-seq data, showing its performance depends heavily on parameter tuning and outperforms other methods when properly tuned.

The paper identifies shared molecular pathways between ANCA-associated vasculitis and nephrotic syndrome, highlighting JAK-STAT signaling as a potential therapeutic target.

LIMBR reduces batch effects and imputes missing data in time series proteomics by modeling experimental structure and circadian patterns.

This paper reviews matrix factorization applications for systems-level omics analysis, highlighting their ability to uncover biological knowledge from high-dimensional data.

PathCORE-T identifies and visualizes globally co-occurring pathways across transcriptomic datasets by analyzing features from machine learning models.

Deep learning rivals or beats baselines across imaging, health records, genomics and drug discovery but is not yet transformative; scarce labels, privacy and opacity remain hurdles.

ADAGE signature analysis identifies perturbed biological processes by learning gene signatures from public data, outperforming curated gene sets in revealing novel processes in Pseudomonas aeruginosa.

The paper develops a method combining functional network community detection with pathway enrichment to reduce false positives in complex gene sets.

This paper trains a variational autoencoder on TCGA RNA-seq data to model cancer gene expression and identify biologically relevant latent features.