Improving Quantitative Evidence for Genetic Tests: Carlos Araya, Invitae

Warning: the first part of this story can sound quite typical. Three co-founders with backgrounds in genomics and AI found a Stanford spinout. Their goal: to bring the tools of AI and computational modeling to unlock the medical secrets of the genome and deliver those to patients. They call this company, Jungla—Spanish for “jungle”—naming not only their adventure, but the whole problem. Then things get interesting.

For their first beta testing, they go around the genetic testing industry looking for partners. Basically what they hear back, is we don’t need that. However, they don’t hear the real reason why these companies don’t need their product. The real reason is because it offers a level of quality and standardization for genetic testing that is not currently required by the industry. In fact it shows “just how vulnerable the industry is.”

But one company—one of the bigger ones, one of the more innovate in the field—said yes, we want to partner. That company was Invitae. And a year later, Invitae would buy out Jungla.

Our guest today is Carlos Araya, the CEO and co-founder of Jungla and now acting as Head of Functional Modeling at Invitae.

It’s the story of one scientist's journey from academia, driven because of a personal family loss, into entrepreneurship, and the story of one company and then one bigger company each saying, “everything can be done better."

Has Computational Modeling for Cancer Genomics "Arrived?" with Shirley Liu, Dana Farber

It’s the question of the moment Are we living in the age of AI? Or is it still just hype?

When it comes to the latest research in immuno therapy, computational modeling is helping to answer key open questions, such as which patients might respond to which drugs.

"If you were to ask me last year about deep learning, I would probably say, aaah, most of the algorithms that are published are not really answering the important questions yet. But I think this year I am converted. We are starting to use deep learning, and we are starting to see interesting results.”

Shirley Liu is a Professor of Biostatistics at Harvard and the Co-Director of the Center for Functional Cancer Epigenetics at the Dana Farber Cancer Institute. Her lab has very recently put out three algorithms, TRUST, TIMER, and TIDE which represent some very exciting ways that bioinformatics is empowering not only cancer research but treatment decisions.

As a computational biologist, Shirley has found herself highly in demand today as the latest genomic tools such as single cell sequencing generate new amounts of data and as public databases such as TCGA make their rich cohorts available.

In today’s interview, she details these three new algorithms and makes the case that computational modeling has arrived for cancer genomics.