*Ashwin Purohit

Demo-Interactive Presentation – Business Track
Sunday, Sept 30, 2012: 11:55 AM – 12:10 PM – Demo Pavilion

← Back to Medicine X 2012 Proceedings
*Presenting Speaker

The problem
Despite widespread agreement on the clinical utility of family medical histories, there exists no easy-to-use tool for collecting them from, and distributing them to family members. A physician hoping for a portrait familial risk must to extract it, question by question, and is met with shoulder shrugs and wrong information. Even in the rare case of a sterling family disease history, the doctor would have to calculate why, for example, the patient’s second cousin’s first heart attack, at age 40, matters to the patient — and by how much.

23andMe solves this problem for physicians with a new, free, online family tree for patients to document their family’s history of disease. 23andMe calculates how everyone in the family, given previous diagnoses, is placed at risk for a host of diseases. The tree is “multiplayer”, so members from the same family can log-in to fill in the gaps about who really had what disease, in the process creating a canonical version of their family’s history of disease. 23andMe’s calculations are simple and informative: a doctor scans a print-out of the patient’s tree, dotted with color-coded diseases, and knows the what and why of a patient’s risk: for example, “Patient is at substantially increased risk for Type 2 Diabetes because of his uncle’s early-onset of Type 2 Diabetes at age 40”.

A demo
Check out our live, beta interface: http://www.screenr.com/fdrs, which we’re constantly improving.

Our science
For many conditions, such as heart disease, stroke, and colorectal cancer, our tree calculates if a family member is at typical risk, moderately-increased risk, or substantially-increased risk of having the condition. We base risk calculation for eight common diseases on a paper by CDC scientists (Scheuner et al. (1997)), and extend it to 26 other conditions with incidence data curated from the literature. To determine a family member’s risk for a disease, we look at how closely related they are to afflicted members, and at what age those members were diagnosed.

Future work: More data, more value

Most of our 10,000 beta users have been genotyped, so we could auto-construct a family tree based on shared identical-by-descent (IBD) segments between members. We could let users visually trace, through the generations, Mendelian inheritance for traits such as immunity to poison ivy, and probabilistically for more complex traits such as eye color.

Using our rules-engine, we can suggest that an individual try a kale recipe given their familial risk for type 2 diabetes. Or, from what we know of their genetics, ask them to tell their physician about their increased sensitivity to a blood-thinning drug like Warfarin. We could send reminders to an aging patient to get a colonoscopy more frequently than once every 10 years, given his family’s history of colon cancer.

Our tree can be the nerve center of a patient’s health: a network of family disease, genetic, and phenotypic information rolled into a sensible report for doctors and patients alike.