Exploiting sparse statistics for a sequence-based prediction of the effect of mutations

Research output: Contribution to journalArticlepeer-review

1 Scopus citations


Recent work showed that there is a significant difference between the statistics of amino acid triplets and quadruplets in sequences of folded proteins and randomly generated sequences. These statistics were used to assign a score to each sequence and make a prediction whether a sequence is likely to fold. The present paper extends the statistics to higher multiplets and suggests a way to handle the treatment of multiplets that were not found in the set of folded proteins. In particular, foldability predictions were done along the line of the previous work using pentuplet statistics and a way was found to combine the quadruplet and pentuplets statistics to improve the foldability predictions. A different, simpler, score was defined for hextuplets and heptuplets and were used to predict the direction of stability change of a protein upon mutation. With the best score combination the accuracy of the prediction was 73.4%.

Original languageEnglish
Article number214
Issue number10
StatePublished - 2019


  • Foldability
  • Mutation
  • Protein sequence
  • Residue n-tuple


Dive into the research topics of 'Exploiting sparse statistics for a sequence-based prediction of the effect of mutations'. Together they form a unique fingerprint.

Cite this