A review of ensemble methods for de novo motif discovery in ChIP-Seq data

De novo motif discovery is a difficult computational task. Historically, dedicated algorithms always reported a high percentage of false positives. Their performance did not improve considerably even after they adapted to handle large amounts of chromatin immunoprecipitation sequencing (ChIP-Seq) data. Several studies have advocated aggregating complementary algorithms, combining their predictions to increase the accuracy of the results. This led to the development of ensemble methods. To form a better view on modern ensembles, we review all compound tools designed for ChIP-Seq. After a brief introduction to basic algorithms and early ensembles, we describe the most recent tools. We highlight their limitations and strengths by presenting their architecture, the input options and their output. To provide guidance for next-generation sequencing practitioners, we observe the differences and similarities between them. Last but not least, we identify and recommend several features to be implemented by any novel ensemble algorithm.