Computing the P-value of the information content from an alignment of multiple sequences

The efficient and accurate computation of P-values is an essential requirement for motif-finding and alignment tools. We show that the approximation algorithms used in two popular motif-finding programs, MEME and Consensus, can fail to accurately compute the P-value. Results: We present two new algorithms: one for the evaluation of the P-values of a range of motif scores, and a faster one for the evaluation of the P- value of a single motif score. Both exhibit more reliability than existing algorithms, and the latter algorithm is comparable in speed to the fastest existing.