Candidate Explorer (CE) assesses the likelihood that a putative mutation is causative for an observed phenotype; specifically, the likelihood of recapitulating the phenotype in mice with an independently generated targeted mutation in the same gene. Screening data (phenotype measurements), mutation data, gene data, and linkage data are input to CE, which employs a supervised machine learning algorithm and approximately 60 features of the data to output both a numerical score (ML Score) and a categorical score (Candidate Status) for each mutation-phenotype association. Published information about gene function is not employed in the assessment; thus the CE assessment is purely genetics-based. CE is trained daily with a cumulative dataset currently containing 1,339 verified and 1,820 excluded mutation-phenotype associations encompassing 458 genes targeted using CRISPR/Cas9 and retested in phenotypic screens. The performance of CE (accuracy, precision, and sensitivity/recall) for each categorical score is displayed for the candidate list retrieved in a particular search. Overall, CE achieves 86% accuracy, 82% precision, and 86% sensitivity in predicting verification for mutations rated “good” or “excellent;” such a record has not been attained by a human analyst to date.
Linkage data obtained through screening will be released in phases according to phenotype. Blood cell flow cytometry screening data are now available for search using CE. These data include unpublished, novel mutation-phenotype associations, with detailed pertinent mutation and phenotype information.
Mutation-phenotype associations and their ML Score and Candidate Status are returned by searching the database by phenotypic screen.
Alternatively, searching with default parameters and “screen name” field empty will return the full list of mutation-phenotype associations. Filter the list for specific genes, effect magnitudes, P values of association, etc. by entering those values in the boxes under the column headings.
Links to the CE search page and current list of available phenotypic screens are in the tabs at the top of this page. Please mouse over the adjacent info icon () for an explanation of that parameter.