Metaanalyses - hms-dbmi/RaMeDiES GitHub Wiki

Our statistical models operate at the level of "mutational targets" which are computed from individual-level variant data. Although a genomic variant has only one mutational target with respect to a particular background (e.g., all intronic SNV variants in a specific gene), the same mutational target can result from several different variants. This means that once mutational targets are computed for a set of variants, information about the individual variants is lost and cannot be accurately extracted from the mutational targets. This means that the intermediate mutational target files computed here can be freely shared to enable cross-cohort analyses!

What are mutational targets?

Intuitively, a "mutational target" is the sum of mutation rates for a set of variants in a gene. You can compute the mutational target of an entire gene as the sum of all mutation rates for all possible variants in that gene. You can also consider the mutational target of subsets of variants in a gene, such as all coding SNVs, all intronic indels, or all intronic SNVs with a variant functionality score above some threshold.

Mutational targets cannot be used to identify specific variants!

Once the mutational target for a specific variant is computed (based on that variant's gene, type, and/or functionality score), the original variant cannot be retrieved. This means that once mutational targets are computed for individual-level patient data, the resulting intermediate mutational target files no longer contain any patient information.

Why do we care?

Files containing mutational targets can be shared without violating restrictions placed on sharing variant-level or patient-level data to enable meta-analyses with other rare disease cohorts!