This story is part of a series on A.I. in healthcare.
Onno Faber was a member of Silicon Valley’s happy breed of tech startup founders when he was diagnosed with a rare genetic condition that can come with dire health damage, but few treatments.
Faber responded with entrepreneurial zeal, exploring whether Silicon Valley’s mastery of algorithms might help root out and defeat the threatening quirks in his genetic code. Without any ready-made solutions on hand from big drug companies and their established research teams, Faber started to recruit individuals to his cause.
The results of Faber’s crusade so far demonstrate a trait Silicon Valley has in common with living things—a startling talent for self-organization.
Faber’s grassroots initiative has led to a signal event this weekend—an AI Genomics Hackathon involving hundreds of artificial intelligence engineers and life sciences researchers, hosted by Google, which is providing $150,000 worth of processing power and a site for the mass collaboration at Google Launchpad in San Francisco.
“People are coming in from Boston, New York, Minnesota,” says Pete Kane, a co-organizer of the hackathon. In addition to 150 people on-site, a similar number will take part remotely online.
Starting Friday, June 23, the volunteer experts will combine their skills to analyze a dataset that includes a high-quality whole genome sequence of Faber’s genome, and the sequence of a tumor he developed as a result of a disorder diagnosed as neurofibromatosis type 2 (NF2). For comparison with Faber’s DNA, the dataset also includes the genome sequence of Faber’s brother, who is apparently unaffected by the disease.
Depending on the form of NF2 disease, tumors can arise in the skin, spine, eyes, or other organs, though they may be non-cancerous. The genetic disorder can cause blindness, deafness, and other impairments, such as mental retardation.
The hackathon is co-hosted by the NF2 Project, a patient organization Faber founded to encourage others diagnosed with neurofibromatosis type 2 to share their DNA sequences for research purposes. Faber’s contact with Kane became another catalyst for the expansion of his grassroots network, and of the weekend of AI-enhanced research on NF2. Kane leads Silicon Valley Artificial Intelligence (SVAI), a self-organized community of AI enthusiasts focused on the use of machine learning and other computational tools in the life sciences. Faber’s quest came at just the right time for the AI group, which wanted to go beyond meetings and discussions with fellow wonks, and “enter some mode of production,” Kane says.
“They wanted some way to collaborate with each other on some open source projects,” Kane says.
Kane, adding to the hackathon’s organizing team, recruited Laura Montoya, founder and CEO of the grassroots AI career accelerator Accel.ai, which helps people chart a path toward jobs in the field, and is developing workshops and longer educational sequences to train students in relevant programming languages. Accel.ai is the third group hosting the hackathon, along with NF2 Project and SVAI.
Montoya, who has an educational background in biology, machine learning, and deep learning, says the various data analytics tools dubbed “AI” are already used in biology and many other fields. But interest in AI is now spiking because the expansion of processing power through GPUs (graphics processing units) has lifted the former limits on computational resources, making it easier to tackle more ambitious projects.
The scale of the Silicon Valley groups’ planned collaboration on Faber’s challenge took another upward leap when Kane contacted Ben Busby, a genomics outreach coordinator for the National Center for Biotechnology Information who has been organizing hackathons for the agency since early 2015. NCBI is a unit under the National Institutes of Health, the premier funding institution for biomedical research at U.S. universities, and home to thousands of in-house government researchers as well. With Busby on board, NCBI became a community partner with the AI Genomics Hackathon focused on NF2.
While the NIH is at the heart of the national biomedical research establishment, it has been using the highly informal hackathons to recruit outside IT experts to collaborate with NIH staff scientists, and also to generally build a community of data scientists who have gotten hooked on problems in biology. NCBI hosts some of the hackathons, but it also invites outside groups to propose topics and locations for week-long or weekend collaborative geek-fests.
As a bonus from over a dozen of these high-energy meetings, ad hoc teams have produced about 100 open source software products that any researcher can use to assist their work in genomics. The products range from a program that helps users tap into federal agency databases to another tool that helps researchers compare structural variants in a genome sample to those on public databases.
The software is available on GitHub, at the NCBI Hackathons page.
Research publications about the hackathon findings also have their own channel on the U.K.-based open research publishing site F1000Research, a sort of outsider outlet that flips the usual slow procedure required for publication in academic journals. Researchers can get a paper published in about seven days, and peer review comes afterwards, as other experts weigh in with online comments.
Busby will be fresh from hosting another hackathon at the New York Genome Center this week when he comes to the Bay Area to be an emcee and mentor at the AI Genomics Hackathon.
Kane says the organizers will offer hackathon participants some suggested approaches to the work, proposed by Busby as well as experts at U.C. Berkeley and Stanford. But they won’t set strict parameters for the “open-ended, free-form” research revels, where experts can organize themselves into their own teams.
One possible approach will be to look for signs that neurofibromatosis type 2 is associated with biomolecules that are already the targets of existing treatments for cancer, Kane says. That might lead to possible NF2 clinical trials of drugs already on the market.
Another ambitious possibility is to create a new drug using the open source Python library DeepChem, a tool created to ease the use of machine learning in drug development. The creator and lead developer of DeepChem, Stanford computer science PhD student Bharath Ramsundar, will be one of the mentors and judges at the hackathon.
Mentors will circulate among the working groups, and the organizers will dole out processing power from Google in $500 increments, Kane says. He says he’ll be particularly excited to see what results emerge from the teams that keep coming back for more processing units—a possible sign that they’re onto something.
The teams will present their results on Sunday, and the judges will evaluate them based on several criteria, such as their usefulness to accelerate further research in genomics, and their potential to lead to viable products on the market.
But the work that results from the three-day hackathon will be free for use by all comers—universities, companies, and participants in the next hackathons.
“We want everything that comes out of this hack to be open source,” Kane says. “Everything officially submitted will be publicly available.”