Google Launches DeepVariant AI Tool for Genomic Data Interpretation, Available on Google Cloud

Recently, VCBeat learned thatGoogle has released a tool called DeepVariant, which leverages state-of-the-art artificial intelligence technologies to construct a more accurate representation of the human genome from sequencing data.

DeepVariant helps convert high-throughput sequencing reads into images of the whole genome and automatically identifies small insertion and deletion mutations and single-base-pair mutations in sequencing data.。

In the new century, the widespread application of high-throughput sequencing has made genome sequencing remarkably accessible. However, the data generated by gene sequencing provide only a limited and error-prone snapshot of the whole genome. For scientists, it is highly challenging to distinguish small mutations arising from random errors during the sequencing process, particularly those located in repetitive regions of the genome. Yet these mutations may be directly implicated in carcinogenesis.

Currently, there are many tools available on the market for interpreting these data, such as GATK, VarDict, and FreeBayes. These software tools often employ simpler statistical and machine learning methods to identify mutations by attempting to exclude erroneous reads.

Brad Chapman, a research scientist at the Harvard T.H. Chan School of Public Health, stated, “One of the challenges lies in the difficult regions of the genome, where each tool has its own advantages and disadvantages. None are perfect. These challenging regions are becoming increasingly important for clinical sequencing, and the more methods available for interpretation, the better.”

DeepVariant was developed by researchers from the Google Brain team, which focuses on the development and application of AI technologies. Verily, another Alphabet subsidiary, is responsible for life sciences.

In a project called Genome in a Bottle (GIAB), the team collected millions of high-throughput reads and whole-genome sequencing data. The GIAB project is a public-private partnership aimed at promoting genomic sequencing tools and technologies. They input the data into deep learning systems and carefully adjusted the model parameters until the system could accurately interpret genetic data.

Last year, DeepVariant won first place in the Precision FDA Truth Challenge, a competition hosted by the FDA to promote more accurate genomic sequencing.

“Deep Genomics CEO Brendan Frey stated, ‘The success of DeepVariant is significant, as it demonstrates that in the field of genomics, deep learning can be used to automatically train systems that outperform complex, manually engineered ones.’”

The release of DeepVariant marks a new milestone in the application of machine learning to genomics.

Deep Genomics is one of the few companies attempting to use AI methods, such as deep learning, to identify disease-causing genes and discover potential drug therapies. (VCBeat has learned that Xiong Huiyuan, Chief Scientist at Infervision, is a co-founder of this company.)

Frey stated, “Artificial intelligence will continue to advance, going beyond merely assisting in the sequencing of genomic data. The core challenge in the development of targeted therapies currently lies in our inability to accurately correlate genetic variants with disease mechanisms and leverage this knowledge to rapidly identify life-saving treatments.” VCBeat has learned from multiple experts in China that the rapid development of AI technology will uncover previously unknown relationships between diseases and their causes, thereby expanding our understanding of disease.

DeepVariant is also available to users on the Google Cloud Platform. Google and its competitors are integrating machine learning capabilities into their cloud platforms to attract anyone seeking to leverage the latest AI technologies.

Overall, AI will help the pharmaceutical sector achieve significant leaps in many areas in the coming years. AI also presents numerous opportunities to extract diverse types of medical data from images or medical records, such as predicting diseases that human physicians might otherwise miss.

Genomic medicine represents a significant opportunity, as the scale and complexity of data are unprecedented. “For the first time in history, our ability to measure and manipulate biology far exceeds our ability to understand it,” said Frey. “AI is the only technology capable of interpreting and processing this vast amount of data. This will fundamentally transform the future of medicine.”

Source: MIT Technology Review

Original article link: https://www.technologyreview.com/s/609647/google-has-released-an-ai-tool-that-makes-sense-of-your-genome/