PhyloCSF

PhyloCSF identifies protein-coding regions in multi-species nucleotide sequence alignments by analyzing evolutionary signatures characteristic of conserved coding sequences.

Key Features:

• Codon Substitution Frequencies (CSF): Quantifies codon substitution patterns to distinguish coding from non-coding sequence based on Codon Substitution Frequencies.
• Substitution-pattern analysis: Evaluates high frequencies of synonymous codon substitutions and conservative amino acid changes and low frequencies of missense and nonsense substitutions.
• Multi-species alignments: Operates on multi-species nucleotide sequence alignments to detect conserved coding signatures across taxa.
• Whole-genome prediction tracks: Produces genome-wide PhyloCSF prediction tracks for species including human, mouse, chicken, fly, worm, and mosquito.
• Machine learning integration: Integrates machine learning to predict and prioritize novel conserved protein-coding regions and high-scoring candidates for downstream curation.
• Annotation evidence support: Supports interpretation with comparative annotation and evidence types such as single-nucleotide variant patterns indicative of purifying selection, exon splicing from GENCODE transcripts, and mass spectrometry translation evidence.

Scientific Applications:

• Transcript classification: Differentiates protein-coding and non-coding RNAs among novel transcript models from high-throughput transcriptome sequencing.
• Gene annotation expansion: Identified over 1,000 high-scoring human PhyloCSF regions, leading to the confident addition of 144 new conserved protein-coding genes to the GENCODE gene set and discovery of additional coding regions in 236 previously annotated protein-coding genes and 169 pseudogenes.
• Comparative annotation of ORFs: Enables comparative annotation across vertebrate genomes to eliminate spurious open reading frames (ORFs) and distinguish coding regions from pseudogenes.
• Refinement of disease-associated loci: Reclassified 118 GWAS variants previously annotated as non-coding as protein-altering, informing interpretation of genetic contributions to disease.

Methodology:

Analyzes evolutionary signatures in multi-species nucleotide sequence alignments using Codon Substitution Frequencies (CSF) by quantifying synonymous, conservative amino acid, missense, and nonsense substitution patterns, generates whole-genome PhyloCSF prediction tracks, and integrates machine learning to predict novel conserved protein-coding regions.