Table 1.
Some phenotype-based variant prediction tools.
| Tool | Concept | Authors benchmarks | References and software |
|---|---|---|---|
| Exomiser (hiPHIVE) (human/interactome-PHenotypic Interpretation of Variants in Exomes) | Integrated phenotypic and interactome analysis using model organisms (mouse, zebrafish) and human clinical data along with protein-protein interaction network data. Focussed on finding new disease genes. | Known disease-gene associations the top hit in 97 % of simulated exomes. | 1,29,30 http://www.sanger.ac.uk/science/tools/exomiser |
| eXtasy | Integrates predicted impact of variants with haploinsufficiency and phenotype-specific gene prioritisation. Uses random forest learning trained on the Human Gene Mutation Database (HGMD16) | Outperforms classical deleteriousness scores (PolyPhen, SIFT, MutationTaster). | 13 http://extasy.esat.kuleuven.be/ |
| OMIM Explorer | Reduces high dimensional phenotypic and genotypic data using semantic similarity and multidimensional scaling. Interface can be used to convert clinical notes to HPO terms. | Clinical variants given median rank of 2, causal variants in top 1% of candidates (47 cases). Outperformed Phen-Gen, eXtasy, and Exomiser (hiPHIVE) for clinical variants. | 17 http://omimexplorer.research.bcm.edu:3838/omim_explorer/ |
| OVA “Ontology Variant Analysis” | Integrates human and model organism phenotypes, functional annotations, curated pathways, cellular localizations and anatomical terms using supervised learning. Exploits multiple ontologies and experimental interaction data23. | Outperformed ExomeWalker31 in benchmarking with 150 exomes. True disease gene ranked first in 20% on cases. | 18 http://dna2.leeds.ac.uk:8080/OVA/index.jsp |
| Phen-Gen | Semantic matching of symptoms against disorder database following Phenomizer14. Functionally related genes recognised through random walk algorithm. Variants classified using conservation and predicted functionality scores. Phenotypic and genotypic evidence combined in Bayesian framework. | Causal coding variants ranked first in 88% of cases (simulation) and in 8 of 11 patient samples. Outperformed VAAST, eXtasy and Phevor by 13–58% and PHIVE by 13–16%. | 24 http://phen-gen.org/ |
| PhenIX (Phenotypic interpretation of eXomes) | Interrogates only known Mendelian genes and uses semantic similarity matching in Phenomizer14. Uses MutationTaster, Polyphen2 and SIFT to predict pathogenicity. | Tests on 52 patient samples with known mutations correct gene achieved mean rank of 2.1 | 11 http://compbio.charite.de/PhenIX/ |
| Phevor “Phenotype driven variant ontological re-ranking tool” | Uses ontologies to re-prioritise candidates identified by other variant prioritisation tools such as SIFT, PhastCons and VAAST to identify alleles not previously linked to disease. | Improved performance of tools such as SIFT and VAAST. | 26 http://weatherby.genetics.utah.edu/cgi-bin/Phevor/PhevorWeb.html |