The algorithm behind the scores is complex and proprietary, but the basic logic is as follows.
First, there are potentially two types of genetic variations present for a single individual: those that contribute to the strength of a trait and those that decrease the strength of a trait. For example, within a single person there may be genetic variations that increase the likelihood that a person may be predisposed to obesity and there may be others that decrease that likelihood. We look at each genetic variation, determine its influence on the trait, weigh the importance given its potential role in critical metabolic pathways and enzymatic reactions, explore whether it is co-occurring with other variations that we expect to see if there is a higher risk and calculate a net likelihood score for the individual. Then we look at population data and rank the person based on where they are in terms of likelihood of predisposition compared to the rest of the population. The population percentile score shows the percentage of people who have less likelihood than the individual to be predisposed genetically to a trait.