Google releases an AI tool to make more clear human genome

Google announced Monday an open source version of DeepVariant, the artificial intelligence tool that last year earned the highest accuracy rating at the precisionFDA’s Truth Challenge.

The open source tool comes as academic medical centers, hospitals, insurance companies and other healthcare organizations are gearing up for if not already embarking on artificial intelligence, cognitive computing and machine learning as well as precision medicine and the genomic sequencing that entails.

DeepVariant helps turn high-throughput sequencing readouts into a picture of a full genome. It automatically identifies small insertion and deletion mutations and single-base-pair mutations in sequencing data.

High-throughput sequencing became widely available in the 2000s and has made genome sequencing more accessible. But the data produced using such systems has offered only a limited, error-prone snapshot of a full genome. It is typically challenging for scientists to distinguish small mutations from random errors generated during the sequencing process, especially in repetitive portions of a genome. These mutations may be directly relevant to diseases such as cancer.

Google Brain Team members Mark DePristo and Ryan Poplin explained that by open sourcing DeepVariant, which was created with Google’s Verily company, they hope to encourage its use and collaboration.

“To further this goal, we partnered with Google Cloud Platform to deploy DeepVariant workflows on GCP, available today, in configurations optimized for low-cost and fast turnarounds using scalable GCP technologies like the Pipelines API,” they wrote in a Google blog. “This paired set of releases provides a smooth ramp for users to explore and evaluate the capabilities of DeepVariant in their current compute environment while providing a cloud-based solution to satisfy the needs of even the largest genomics datasets.”

To build DeepVariant, the Google Brain team started with reference genomes from the Genome in a Bottle Consortium to create tens of millions of examples for encoding high-throughput sequencing data. The next step was training a Google TensorFlow classification model to distinguish between the true genome and the experimental data, DePristo and Poplin said. They added that DeepVariant won the highest accuracy rate in the FDA’s precisionFDA Truth Challenge in 2016 and, since then, Google has cut the tool’s error rate by more than 50 percent.

Microsoft, for its part, launched a new U.K-based unit focusing on AI for healthcare in September of 2017 after aligning with UPMC in a strategic partnership to advance artificial intelligence and released new products including HealthVault Insights, Microsoft Genomics, a chatbot and Project InnerEye for radiotherapy.

Google said DeepVariant is part of its overarching goal to apply its technologies to healthcare and other scientific disciplines and ultimately make the tools and data accessible to a wide range of people.

“DeepVariant is the first of what we hope will be many contributions that leverage Google's computing infrastructure and machine learning to both better understand the genome and to provide deep learning-based genomics tools to the community,” DePristo and Poplin said. 

The future of artificial intelligence has become a divisive issue in Silicon Valley. Some tech executives side with Facebook CEO Mark Zuckerberg, who has optimistic views of the technology, while others have been swayed by Tesla CEO Elon Musk’s darker predictions for machine learning.