NVIDIA develops AI to speed up human genome analysis

Scientists from NVIDIA and Harvard University have developed the AtacWorks toolkit for deep learning, designed to reduce the cost and time required for experiments with rare and single cells.

AtacWorks, DL toolkit for epigenomics research out in @NatureComms, brings down cost and time needed for experiments:https://t.co/f4aCca8Msg @NVIDIAHealth
Congrats @lal_avantika @z_chiang @ivan_bezdomny @fabianamduarte @JD_Buenrostro!

— Johnny Israeli (@johnny_israeli) March 8, 2021

AtacWorks works with ATAC-seq, a well-established method designed to locate open regions in the genome of healthy and diseased cells. These regions are stretches of DNA that are used to identify and regulate certain functions (for example, liver, blood, or skin cells). This part of the genome may give scientists clues about whether a person has Alzheimer’s disease, heart problems, or cancer.

For ATAC-seq, analysis typically requires tens of thousands of cells. AtacWorks can obtain the same results using just a few dozen. The toolkit can perform genome-wide inference in as little as half an hour, whereas conventional methods typically take a little over two days.

The prospect of faster and cheaper genome analysis will be significant for identifying specific mutations or biomarkers that may lead to certain diseases.

Avantika Lal, an NVIDIA researcher and the paper’s lead author, said that existing methods do not allow studying DNA differences in very rare cell types.

«AtacWorks can help not only reduce the costs of collecting chromatin accessibility data, but also unlock new possibilities in drug development,» Lal added.

Earlier in January, MIT presented an artificial intelligence algorithm capable of diagnosing breast cancer from mammography images more accurately than a radiologist.

Follow Forklog news on Facebook!