Biomathematics / Computational Biology Colloquium

Genomic prediction of disease through quasi-Mendelian loci

Speaker: Lawrence Sirovich, Rockefeller University

Location: Warren Weaver Hall 1314

Date: Tuesday, September 11, 2018, 12:30 p.m.


A novel analysis leads to the discovery of near Mendelian (quasi-Mendelian) loci, and algorithms, virtually hypothesis-free, that produce robust genomic prediction of disposition towards disease or wellness, for all seven diseases of the Wellcome Trust WTCCC1 database.

These claims are confirmed in compelling fashion by an in-depth systematic randomized training/testing analyses; and independent verification from application of the machine learning algorithm, Random Forests.

For each disease studied, a modest number of predictive, quasi-Mendelian, alleles are aadequate, roughly O(102).

It is the contention of this presentation that practical disease prediction is possible for all seven diseases of the WTCCC1 database for subjects of European stock, the ancestry of the WTCCC1 cohorts.