- AtacWorks uses Nvidia’s NGC hub of GPU-optimized software, can study the genome within 30 minutes.
- The researchers claim that the testing time and the cost could be eliminated significantly.
Nature Communications is a journal where the co-authors showed that AtacWorks can analyze the entire genome within half an hour. On the other hand, the traditional methods required hours of testing. The maximum number of cells provides a total copy of a person’s DNA where a billion pairs are sardined into the nucleus of the cell. But, when the cell has a requirement, it pulls out a section of the genome and starts the function. These cell types include liver, blood, or skin.
About the technology behind AtacWorks:
Nvidia’s NGC hub of GPU-optimized software provided AtacWorks. It perfectly functions with ATAC-seq which is a technique to detect open areas in the genome in the cells. Harvard professor Jason Buenrostro pioneered this method and he is also one of the co-authors of the paper. ATAC-Seq is responsible to measure the signal intensity at every point in the genome. When the peaks in the results appear, it means that the cells are fewer.
ATAC-Seq requires thousands of cells to receive a clear signal. But surprisingly, AtacWorks can work with just ten cells to produce the same quality of results, according to the co-authors. AtacWorks has been trained on the labeled pairs of matching ATAC-Seq datasets, where one is high-quality and the other noisy. With a downsampled data copy, the model learned to predict a very accurate, high-quality version and could identify peaks in the signal.
With AtacWorks, the researchers could see that it was easy to spot chromatin, which packages long molecules into small, complex structures. Hence, in a noisy sequence of 1 million reads, the model could read the information as accurately as the traditional methods that read a clean dataset with 50 million reads.
What the researchers have to say:
AtacWorks can let scientists and researchers work on a small number of cells. This will reduce the cost of sample collection and sequencing. It will also make the analysis cheaper and quicker. AtacWorks worked out inference in 30 minutes with the Nvidia Tensor Core GPU. Normally, through traditional methods, the process would take 15 hours on a system running on 32 CPU cores.
In the Nature Communications paper, the Harvard researchers apply AtacWorks to a dataset of stem cells that would give out red blood cells and white blood cells. These are rare subtypes that cannot be studied through traditional methods. The team working with AtacWorks took a sample set of only 50 cells which identified the distinct areas of DNA which were associated with the development into white blood cells. It would also separate the sequences which would correlate with the red blood cells.
“With very rare cell types, it’s not possible to study differences in their DNA using existing methods,” Nvidia researcher Avantika Lal, first author on the paper, said. “AtacWorks can help not only drive down the cost of gathering chromatin accessibility data, but also open up new possibilities in drug discovery and diagnostics.”
