A Bayesian Inference Model to Predict Phenotype from Personal Variation

C12508: Bayesian Inference Model to Predict Phenotype from Personal Variation

Novelty:

An inference model that predicts the probability of medically relevant phenotypes from an individuals genomic information.

Value Proposition:

Genetic testing is increasingly used as a diagnosis, linking genomic data to specific phenotypic characteristics; however, due to patient variability, it is challenging to develop a model to predict the association between genetic information and phenotype on an individual level, for personalized testing. This model can be used to predict individual phenotypes based on an individuals genetic information. The model outputs the probability that the individual has the phenotype(s) of interest, providing a novel, personalized approach to clinical genetics. Additional advantages of this model include:

Technical Details:

Johns Hopkins researchers have developed a Bayesian inference model to predict individual phenotypes based on personal genetic information. A Bayesian network was built for each predicted phenotype. Model parameters were estimated from population statistics, including the prevalence and the heritability of each phenotype and predicted functional impact of variants using the existing Variant Effect Scoring Tool. Bernoulli random variables are used to represent unknown mechanisms affecting the penetrance of affected variants. An individuals variome is used as the model input. The model computes the probability that the individual has each assessed phenotype and calculates the weighted Bernoulli likelihoods.

Looking for Partners:

To develop and commercialize the technology for personalized genetic testing.

Stage of Development:

Demonstrated proof of concept by matching 293 phenotypic profiles to the variomes of 77 individuals

Data Availability:

Prototype