Challenges in adjusting scoring matrices when comparing functional motifs with non-standard compositions

Methods for scoring matrix adjustment decrease the significance of biased residues to better detect homology between protein sequences. This is because non-homologous proteins often contain fragments with non-standard compositions that are strikingly similar to each other. However, these fragments are also functionally important in proteins and are receiving an increasing attention from the scientific community. In this study, we described why the gold standard method for scoring matrix adjustment is unable to emphasise frequent amino acids. Further, we used BLAST to align collagen-like domains with and without the scoring matrix adjustment and compared the results. We found that the scoring matrices were adjusted in the opposite direction to the optimal state. Therefore, turning off the adjustment improved alignment quality of collagen-like domains, but scoring matrices still need refinement. This study provides a detailed analysis of why the gold standard method fails, and opens doors for new methods to adjust scoring matrices for functional motifs with non-standard compositions.