Thomas P. Quinn

This paper introduces CoDaCoRe, a deep learning method for identifying sparse log-ratio biomarkers from high-throughput sequencing data, achieving faster performance and better accuracy than existing methods.

This paper introduces a neural network model called TATA to learn distance-dependent motif interactions in RNA splicing using self-attention and Toeplitz max pooling, and applies it to identify cis-regulatory elements interacting with splice sites.

This paper introduces PAN, a method to create sample-specific gene networks using annotations, improving breast cancer relapse prediction compared to existing methods.

The paper predicts gene-drug responses for individual patients by integrating single-cell expression data with drug-response databases using a dual-channel random walk method.

BrainGENIE predicts brain-tissue gene expression from peripheral blood transcriptomes, outperforming existing methods in accuracy and biological relevance.

Representing drugs as molecular graphs, GraphDTA lowers drug-target affinity prediction error to MSE 0.229 on Davis and 0.139 on Kiba versus deep learning baselines.

This paper introduces a data-driven method for dimensionality reduction of compositional data using amalgamation, achieving performance comparable to state-of-the-art methods while producing interpretable features.

The paper introduces HyperXPair, an interpretable neural model that learns distance-dependent motif interactions for transcription initiation and RNA splicing.

The paper introduces a sparse neural encoder-decoder network to predict gut metabolites from microbiome data, showing improved accuracy and interpretability over traditional methods.

The paper argues that simple linear statistics can outperform the mmvec neural network in analyzing microbe-metabolite correlations.

This paper provides a guide for applying compositional data analysis to NGS count data, emphasizing log-ratio transformations and addressing limitations of conventional normalization methods.

This paper introduces an anomaly detection approach to classify cancer using only healthy tissue transcriptomes, showing improved accuracy with more normal samples.

This paper improves neuropsychiatric condition classification by using gene ontology terms as features, enhancing performance, stability, and interpretability of classifiers.

This paper explores using balances to engineer features for classifying health biomarkers and benchmarks transformation methods across different biomarker types.

This study identifies sex-dependent and -independent biomarkers in ASD using post-mortem brain RNA-Seq and blood microarray data, highlighting sex-specific gene expression differences and pathway dysregulation.

The paper benchmarks ALDEx2, a log-ratio transformation method, against conventional RNA-Seq differential expression tools, showing it has high precision and comparable performance on real data.

This paper reviews compositional data analysis principles and discusses their application to sequencing data, highlighting challenges and future directions.

This paper introduces a normalization-free method for detecting differential gene expression by analyzing changes in gene ratio proportionality between sample groups.

The paper introduces an R package for analyzing compositional data by implementing two proportionality measures and discussing their application to genomic data.