The goal of KD4v server is to study and understand the links between the structural impact of a mutation and the human disease phenotype. We applied Inductive Logic Programming (ILP) approach to the missense variants mapped to 3D protein structure and annotated by MSV3d database

This server provides two main complementary services:

(i) A knowledge base consists ILP rules characterizing deleterious mutations that can be biologically interpreted interpretable by biologists and thus can guide human expert discovery of new correlations between mutation, sequence/3D structure and phenotypic severity. These rules are linked to richer relevant knowledgebase including: (alignment information, protein structure, protein interaction, gene expression, omim, etc);

(ii) Mutation prediction based on ILP rules. ILP approach can be effectively used for mutation effect prediction with performances similar to the most widely used methods: SIFT, Polyphen-2.


Resources statistics in KD4v server
  • 10713 human proteins with the 3D models available on our server. The 3D models can be constructed since the query and template sequences share at least 50% identity.
  • User could discovery and predict more than 32.000.000 new missense variants (all position of 3D mapped and all possible amino acid replacement in protein).
  • Background knowledge with 16 predicates corresponding to the mutation data and the Prolog build-in comparison predicates.


  • Please cite: Tien-Dao Luu, Alin Rusu, Vincent Walter, Benjamin Linard, Laetitia Poidevin, Raymond Ripp, Luc Moulinier, Jean Muller, Wolfgang Raffelsberger, Nicolas Wicker, Odile Lecompte, Julie D. Thompson, Olivier Poch, and Hoan Nguyen (2012). KD4v: Comprehensible Knowledge Discovery System For Missense Variant. Nucleic Acids Res
    doi:10.1093/nar/gks474