How NVIDIA is helping researchers and health officials use AI to progress genetics.
By Dereck Severino
As the field of genetics grows and expands rapidly, so is technology and AI. Genome sequencing has been very reliant on advanced computer systems in order to rapidly acquire and read all the genetic data. The improvement of genome sequencing is highly reliant on the improvement of scientists’ genome sequencing hardware and software. With the death of Moore’s law (the law that the number of transistors in an integrated circuit doubles about every two years), chip manufacturers have had to turn to new ways of improving their hardware. AI is quickly being implemented in bioinformatics by companies such as NVIDIA, a GPU manufacturer. NVIDIA has been a leader in implementing AI systems into their products.
Artificial intelligence has sparked a new revolution in bioinformatics. The cost of sequencing a human’s genome continues to decrease while the storage capacity required continues to increase. AI deep learning is being implemented along with neural networks in chips in order to interpret the visual data. The AI is capable of interpreting image and signal data generated by instruments and inferring the 3 billion nucleotide pairs of the human genome. Based on this, it can also quickly identify patterns and potential mutations in real time.
According to their solution brief, NVIDIA wants to branch AI into computer science with the NVIDIA Clara™ platform. Their goal is to help raise the pillar of precision medicine, medicine that can be given based on a person’s genome and not a statistical average. They say that GPUs along with AI deep learning tools such as DeepVarient can accelerate data analysis making what would have taken a CPU hours into a couple seconds. An example of this can be shown in Long Read Sequencing (LRS). Pacific Biosciences is a leader in the industry of LRS. They’re reaching new heights of sequencing with their “HiFi technology” which is circular consensus sequencing in combination with their deep learning AI model, DeepConsensus. This system in combination with NVIDIA’s powerful GPUs can sequence up to 1300 human genomes per year. Researchers at Stanford were able to achieve the world record for the fastest DNA sequencing technique which was powered by NVIDIA’s GPUs.
NVIDIA’s endeavors are not limited to just sequencing. NVIDIA is also looking to improve downstream analysis for predicting variant pathogenicity and gene expression. NVIDIA is using Large Language Models (LLMs) in order to accomplish this. They are AI models that were taught large datasets of labeled genomic data to accurately predict pathogenic variants and gene expression. NVIDIA also has used their AI to improve Spatial Genomics which is a technique to sequence cells in a tissue, while retaining the spatial information of those cells. NVIDIA were able to improve NanoString’s CosMx Spatial Molecular Imager to get 5–20X faster cell
segmentation in it compared to slower workloads running on CPUs.
NVIDIA is focused on enabling the next wave of genomics by powering biotechnology companies and researchers with AI accelerated hardware. NVIDIA is working with biotech companies to push the boundaries of whole genome sequencing.