The Google-developed DeepVariant bioinformatics tool is being offered by DNAnexus on its cloud-based platform through a pilot program.
Launched last week, DeepVariant is a deep-learning technology designed to call genetic variants from next-generation sequencing data, using deep neural networks.
DeepVariant applies the Inception TensorFlow framework, originally developed to perform image classification. The framework represents information as images similar to what a clinician would look at from a genome browser, then classifies the positions as variant or non-variant.
By cross-applying the technology using a short development time, DNAnexus reasons, DeepVariant can outperform methods that developers have been working on for a decade,
“Deep learning-based data analysis tools have tremendous potential in driving further advances in our understanding of disease biomarkers,” DNAnexus CEO Richard Daly said yesterday in a statement. “DNAnexus is able to host scalable architecture to support these complex technologies.”
Announcing DeepVariant in a December 4 post on the Google Research Blog, Google Brain Genomics leader Mark DePristo, Ph.D., and technical lead Ryan Poplin stated that the bioinformatics tool’s purpose is “to reconstruct the true genome sequence from HTS sequencer data with significantly greater accuracy than previous classical methods.”
Not so, countered Steven Salzberg, Ph.D., Bloomberg Distinguished Professor of Biomedical Engineering, Computer Science, and Biostatistics at Johns Hopkins University, in a commentary published Monday by Forbes, where he is a contributor: “DeepVariant is just a variant caller (as the name implies). This software does not ‘reconstruct the true genome sequence.’ That's just wrong. To reconstruct the sequence, you would need to use a genome assembler, a far more complex algorithm.”
Through an initial pilot program, DeepVariant will be offered to an unspecified “limited number of interested users, with broader access to the tool in the coming months.”
“The true power of DeepVariant lies not in its ability to accurately call variants—the field is mature with solutions to do so,” DNAnexus VP of Science Andrew Carroll, PhD and bioinformatician Naina Thangaraj said in a December 5 post on the company’s blog. “The true power is as a demonstration that with similar thoughtfulness, and some luck, we could rapidly achieve decades of similar progress in fields where the bioinformatics community is just beginning to focus effort,”
DNAnexus’ platform, designed to assist researchers and clinicians accelerate global genomics programs across industries and institutions that include biopharma, bioagriculture, sequencing services, clinical diagnostics, government, and research consortia.
“By making a reference implementation broadly accessible, I believe Google and DNAnexus will accelerate a growing body of research that requires high quality genomic information,” stated Vikram (Vik) Bajaj, Ph.D., managing director at Foresite Capital Management and former CSO at Verily Life Sciences (formerly Google Life Sciences).
The technology behind DeepVariant was initially described in a December 2016 pre-publication paper. An earlier version of DeepVariant won the PrecisionFDA Truth Challenge Award for overall accuracy in calling of single nucleotide polymorphisms (SNPs), following its submission to the DNAnexus-powered PrecisionFDA platform. Since then, the development team has continued to train the system and has decreased the variant calling rate by another 50%.
DeepVariant is being made available to the genomics community as an open-source tool on the Google Cloud Platform.
Founded 2009, DNAnexus is based in Mountain View, CA. The company says its platform is built on Microsoft Azure and Amazon Web Services and is designed to help genome centers migrate current processes and workflows to the cloud.
