BLAST calculates expect values that describe the significance of a match, with a lower expect value indicating a more significant match. Since the 2.2.26+ release, BLAST+ uses an improved method to calculate the statistical significance of protein-protein matches. The new method uses a better finite-size correction (FSC) to improve the accuracy of results. The new FSC calculation approximates the distribution of the lengths of the optimal matches in the query and subject sequences, not just the corresponding means. This improvement is especially important for matches with short sequences, because the older method could underestimate the significance of such a match by many orders of magnitude. An article in BMC Research Notes by Park et al., describes these improvements.
↧