Pseudogenes are the remnants in genomic sequences of genes which are no longer functional. They are frequent in Eukaryotic genomes, and an important resource for comparative genomics. However, pseudogenes are often mis-annotated as functional genes in sequence databases. Current methods for identifying pseudogenes include methods which rely on the presence of stop codons and frameshifts, as well as methods based on the ratio of silent to replacement nucleotide substitution rates (dN/dS). A recent survey concluded that 50\% of human pseudogenes have no detectable truncation in their pseudo-coding regions, indicating that the former methods lack sensitivity. The latter methods have been used to find sets of genes enriched for pseudogenes, but are not specific enough to accurately separate pseudogenes from expressed genes.
Here we introduce a a program called Psueodgene Inferrence from Loss of Constraint (PSILC): novel methods for separating pseudogenes from functional genes. The methods calculate the log-odds score that evolution along the final branch of the gene tree to the query gene has lost versus not lost the domain encoding constraint in favour of protein coding/ neutral nucleotide evolution. Using the manual annotation of chromosome 6, we show that both of these methods result in a more accurate classification of pseudogenes than dN/dS.