The choice of sequence homologs included in multiple sequence alignments has a dramatic impact on evolutionary conservation analysis

Abstract

Supplementary data are available at Bioinformatics online.

Assessing the accuracy of contact predictions in CASP13

“…Even if we consider the single best prediction out of all the groups for each of the 32 targets, we get an overall average F ‐score of 0.24 (the highest single F ‐score achieved by any group and any target is 0.76 in this set). A recent work estimated that random residue‐based prediction results in an average F ‐score of approximately 0.12, with a very sharp normal distribution. Therefore, although an average F ‐score of 0.24 is certainly statistically significant, it is also clear that there is much room to further improve contact predictions.…”

Discovery of receptor‐ligand interfaces in the immunoglobulin superfamily

“…Sequence‐based interface predictions were performed as follows, adapted from previous studies: A given query protein's amino acid sequence was searched through the NCBI “nr” database using jackhmmer 3.1 with a domain‐based e‐value cutoff of 10 −20 and otherwise default parameters, generating a sequence profile typically including several thousand hits. The jackhmmer profile was then subset into 264 alternative MSAs by combinatorially applying three sequence identity filters: the minimum (set at 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, and 60%) and maximum (set at 50%, 70%, 90%, and 99%) sequence identity between query and hits, and the maximum sequence identity (clustering level) among hits (set at 40%, 50%, 60%, 70%, 80%, 90%, 95%, and 99%). The total number of combinations of all parameters is 288 but the minimum and maximum sequence identities of hits to the query have an overlap in the middle range, which reduces the possible number of combinations to 264.…”

“…On the other hand, many machine learning approaches have been developed that combine sequence and structural features to arrive at binding interface predictions . Recent benchmarks suggest that the field of feature‐based binding interface prediction appears to have saturated, as the addition of new properties results in little improvement in performance, and argue that future improvements may be expected from customized predictors that focus on specific classes of proteins …”

“…We present a novel structure‐based interface mapping and filtering approach that we used in combination with our recently‐developed multiple sequence alignment (MSA) selection pipeline, selection of alignment by maximal mutual information (SAMMI) . We tested the algorithms on the set of 22 IgSF proteins with currently‐known trans ‐binding interfaces (Table ), and show that sequence‐based predictions can be used to improve upon structure predictions toward the theoretical limit of binding site prediction . The current limiting factor in achieving this is identifying the optimal set of homologs to include in multiple sequence alignments used in conservation analysis.…”

“…16 On the other hand, many machine learning approaches have been developed that combine sequence and structural features to arrive at binding interface predictions. 20 Recent benchmarks suggest that the field of feature-based binding interface prediction appears to have saturated, as the addition of new properties results in little improvement in performance, 16,21 and argue that future improvements may be expected from customized predictors that focus on specific classes of proteins. 16 In the current work, we combine structural and sequence information to predict the binding interfaces of IgSF proteins ( Figure S1), taking advantage of their structural homogeneity in the context of significant sequence divergence.…”