Abstract
BD-Func (BiDirectional FUNCtional enrichment) is an algorithm that calculates functional enrichment by comparing lists of pre-defined genes that are known to be activated versus inhibited in a pathway or by a regulatory molecule. This paper shows that BD-Func can correctly predict cell line alternations and patient characteristics with accuracy comparable to popular algorithms, with a significantly faster run-time. BD-Func can compare scores for individual samples across multiple groups as well as provide predictive statistics and receiver operating characteristic (ROC) plots to quantify the accuracy of the signature associated with a binary phenotypic variable. BD-Func facilitates collaboration and reproducibility by encouraging users to share novel molecular signatures in the BD-Func discussion group, which is where the novel progesterone receptor and LBH589 signatures from this paper can be found. The novel LBH589 signature presented in this paper also serves as a case study showing how a custom signature using cell line data can accurately predict activity in vivo. This software is available to download at https://sourceforge.net/projects/bdfunc/.