Gene expression and cancer classification

This paper explains why PCA and tensor decomposition work well for feature selection in genomic data by relating them to projection pursuit and empirical P-value thresholds.

The paper introduces TVscript, a tool to filter out transcripts with high intra-condition variation, improving differential expression analysis in multi-species RNA-seq data.

The paper evaluates RNA-seq differential analysis methods across simulated and real data, finding EBSeq and DESeq2 perform best under specific conditions.

This paper identifies major sources of unwanted variation in cancer RNA-seq data and proposes a novel method to remove them using pseudo-replicates of pseudo-samples.

BioTEA is a containerized R/Bioconductor-based tool for differential expression analysis of microarray data, offering reproducibility and flexibility through customizable pipelines.

omicsGAT integrates graph-based learning with attention mechanisms to improve cancer subtype analysis using RNA-seq data.

This study benchmarks eleven feature extraction methods for identifying blood-based transcriptomic biomarkers in cancer diagnosis, finding PLS-DA superior for classification and Ranger for robust feature selection.

This paper introduces a Reproducibility Score to assess biomarker set reproducibility and provides methods to estimate it using univariate hypothesis testing on dichotomous groups.

The paper introduces a method to handle omics data variation across platforms and times, showing its use in three disease risk models for clinical application.

This paper proposes a combined test for feature selection in sparse metaproteomics data, avoiding missing value imputation by addressing both missingness mechanisms and differential intensity.

The paper introduces MIAMI, a method using mutual information to analyze protein co-expression in cancer datasets, finding associations between marker co-expression and clinical outcomes.

The paper improves a tensor decomposition method for integrating gene expression profiles from multiple tissues and identifies disease associations beyond individual tissue analyses.

The paper introduces gmcNet, a graph neural network for WGCNA module clustering, showing improved performance in modularity and differential expression compared to existing methods in Korean native cattle.

The paper develops a sparse negative binomial classifier for RNA-seq data that adjusts for covariates and improves prediction accuracy and feature selection.

The paper identifies ubiquitously expressed genes (UEGs) in humans using a data-driven approach and provides their global expression patterns as a resource for further study.

This study introduces DeepBioGen, a deep generative model for augmenting sequencing profiles to improve prediction model generalizability on unseen data.

The paper introduces a fast and accurate imputation method for miRNA expression data using constrained least squares, outperforming existing methods in speed and accuracy.

MoSBi automates biclustering analysis across multi-omics data to identify robust molecular signatures for stratification and subtyping.

This study evaluates 15 machine learning algorithms and 30 gene expression signatures for prostate cancer prognosis using 1,558 patients, finding survival analysis models outperform binary classifiers and identifying top signatures comparable to commercial panels.

ExhauFS performs exhaustive feature selection for classification and survival regression, demonstrating its utility on cancer datasets.

The paper identifies transcriptome diversity as a major source of systematic bias in RNA-seq data and shows it explains significant variability in gene expression across organisms and datasets.

The paper shows that popular differential expression methods like DESeq2 and edgeR have high false discovery rates in population-level RNA-seq data, recommending the Wilcoxon rank-sum test instead.

The paper benchmarks 50 classification algorithms on 50 gene-expression datasets to evaluate their performance in patient subgroup classification.

This study uses deep learning to classify 37 cancer types from TCGA data, identifying cancer-specific gene expression signatures with >97% accuracy.

ALOHA improves dose-response analysis by using Bayesian splines to model gene responses and cluster co-regulated genes, leading to better identification of biological pathways and benchmark doses.

The paper introduces 2s-LCA, a method for joint analysis of multiple biological data sets, and applies it to human brain development data to identify shared gene expression patterns.

PairGP models longitudinal gene expression data from paired multi-condition studies to identify condition groups with different dynamics.

The paper identifies cancer-related genes by analyzing changes in gene expression variability and distribution, finding distinct sets of genes missed by traditional differential expression analysis.

oCEM detects and analyzes overlapping co-expressed gene modules using decomposition methods and a permutation-based approach to select principal components, outperforming existing techniques.

PanClassif improves cancer classification from single-cell RNA-seq using machine learning by selecting effective genes and applying smoothing and oversampling techniques, achieving better performance than existing methods.