MutPred2 is a machine learning-based method and software package that integrates genetic and molecular data to reason probabilistically about the pathogenicity of amino acid substitutions. This is achieved by providing (1) a general pathogenicity prediction, and (2) a ranked list of specific molecular alterations potentially affecting the phenotype. It is trained on a set of 53,180 pathogenic and 206,946 unlabeled (putatively neutral) variants obtained from the Human Gene Mutation Database (HGMD) , SwissVar , dbSNP  and inter-species pairwise alignment. The MutPred2 model is a bagged ensemble of 30 feed-forward neural networks, each trained on a balanced subset of pathogenic and unlabeled variants.
MutPred2 was developed by Vikas Pejaver at Indiana University, Bloomington, and was a joint project of the Mooney group at the University of Washington and the Radivojac group at Indiana University. The Iakoucheva and Sebat groups at the University of California, San Diego provided additional validation and support. More information on the method and detailed instructions can be seen in the help page.
Pejaver V, Urresti J, Lugo-Martinez J, Pagel KA, Lin GN, Nam H, Mort M, Cooper DN, Sebat J, Iakoucheva LM, Mooney SD, Radivojac P. Inferring the molecular and phenotypic impact of amino acid variants with MutPred2. Nat. Commun. 11, 5918 (2020)
This work is funded by:
NIH R01LM009722 (PI: Mooney)
NIH R01MH105524 (PI: Iakoucheva and Radivojac)
NIH R01MH104766 (PI: Iakoucheva)
NIH R01MH076431 (PI: Sebat)
Indiana University Precision Health Initiative (PI: Radivojac)
NIH K99LM012992 (PI: Pejaver)
The MutPred suite
Beyond MutPred2, several other tools have been developed as part of the MutPred project. They are listed below in chronological order of their development:
MutPred: the proof-of-principle predecessor to MutPred2 that uses a random forest model trained on a smaller training set with fewer features (and predicted molecular mechanisms).
Functional regulatory SNP predictor: an ensemble of decision trees for the prediction of SNPs in regulatory regions that impact gene expression. A new and enhanced method, called RSVP has been developed, and uses an expanded feature set that includes information from ENCODE.
MutPred-Splice: a random forest-based approach to prioritize exonic variants (missense or samesense) which are likely to disrupt pre-mRNA splicing from whole-genome sequencing data sets.
MutPred-LOF: a neural network-based predictor of pathogenic loss-of-function (frameshifting indels and nonsense) variants and their impact on protein structure and function.
MutPred-Indel: a neural network-based predictor of pathogenic non-frameshifting indels and their impact on protein structure and function.
1. Stenson PD, Mort M, Ball EV, Evans K, Hayden M, Heywood S, Hussain M, Phillips AD, Cooper DN. The Human Gene Mutation Database: towards a comprehensive repository of inherited mutation data for medical research, genetic diagnosis and next-generation sequencing studies. Hum Genet (2017)
2. Mottaz A, David FP, Veuthey AL, Yip YL. Easy retrieval of single amino-acid polymorphisms and phenotype information using SwissVar. Bioinformatics (2010) 26(6):851-852
3. Sherry ST, Ward MH, Kholodov M, Baker J, Phan L, Smigielski EM, Sirotkin K. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res (2001) 29(1):308-311