Discover knowledge by Inductive Logic Programming



Knowledge base about human monogenic diseases

Description Set of rules



All human proteins

Description Performance (*) Set of rules
Sensitivity Specificity
1  We used HumVar (of Uniprot) variant set (also used in PolyPhen-2) to train and test KD4v server. We annotated this variant set by SM2PH/MSV3d pipeline and selected only those missense variants that are mapped to 3D protein structures. These missense variants were split into 2 parts, training set (6000 disease-causing mutations and 2000 neutral polymorphisms) and validation set (658 disease-causing mutations and 298 neutral polymorphisms).

The prediction engine of KD4v was built by using the rules learnt on this training set.

74.00 % 
68.00 % 
Download
(in Prolog code)

(*) Performance is reported on the validation set.



Group of genes concerning different genetic disorders

Description Set of rules
1  We gathered all the genes/proteins concerning retinal diseases (379 proteins) including 805 missense variants 3D (659 disease-causing mutations and 146 neutral polymorphisms). This rule set is currently for only analyze.

Download
(in Prolog code)