recount_Bioconductor

recount_Bioconductor provides processed RNA-seq data and phenotype predictions, including RangedSummarizedExperiment objects at gene, exon, and exon-exon junction levels, raw counts, phenotype metadata, sample coverage and mean coverage bigWig URLs, and phenotype annotations trained on TCGA and GTEx to support differential expression and large-scale expression studies of recount2 samples.

Key Features:

• RangedSummarizedExperiment objects: Supplies RangedSummarizedExperiment objects at gene, exon, and exon-exon junction levels for downstream Bioconductor analyses.
• Raw counts and phenotype metadata: Provides raw count matrices together with sample-level phenotype metadata.
• Coverage bigWig files: Includes URLs for sample coverage bigWig files and mean coverage bigWig files specific to each study.
• In silico phenotyping integration: Integrates phenotype prediction methods to infer missing annotations from expression data.
• Training datasets: Leverages well-annotated datasets from TCGA and GTEx as training data for phenotype prediction.
• Large-scale recount2 corpus: Applies to a standardized set of approximately 70,000 RNA-seq samples processed under the recount2 project.
• Phenotype prediction tooling: Encapsulates phenotype prediction methods in the phenopredict R package with predictions accessible through the recount R package.

Scientific Applications:

• Differential expression analysis: Enables differential expression studies using gene-, exon-, or junction-level counts with Bioconductor packages.
• Annotation of missing phenotypes: Provides predicted phenotype annotations (e.g., sex, tissue type, sample source, sequencing strategy) to supplement incomplete public metadata.
• Cross-sample and project selection analyses: Supports selection and comparison of projects or samples based on predicted biological and experimental characteristics.
• Large-scale expression studies: Facilitates population-scale investigations of normal human variation and disease-related expression patterns using recount2 data.

Methodology:

Distributes RangedSummarizedExperiment objects and raw counts, supplies phenotype metadata and bigWig URLs, applies in silico phenotyping methods trained on TCGA and GTEx, and builds and evaluates predictors for biological and experimental phenotypes using gene expression from recount2 samples; phenotype prediction methods are implemented in the phenopredict R package with predictions exposed via the recount R package.