AI in genomics: Progress through innovation

After a person's genomic information is collected, AI and machine learning help analyze the data to determine personalized treatment options. The results can impact pharmacology, oncology, infectious diseases, and many other areas of healthcare. In practice, AI applications in genomics tend to target tasks that are difficult or impractical to complete using human intelligence alone. AI is also useful for tasks that are prone to error with standard statistical approaches, including variant calling, genome annotation, variant classification, and phenotype-to-genotype correspondence.

NetApp customer WuXi NextCODE has created a unique platform specifically to organize, mine, share, and apply genomic data to improve human health. Over the past 20 years, it has amassed the world's largest database of human genome sequences. WuXi NextCODE uses NetApp Cloud Volumes Service to help researchers quickly generate insights from processing unprecedented amounts of genomic data, discovering new ways to address disease. Across four continents, the company's partners can correlate genomic and phenotypic data at unprecedented scale and speed.

In the case of ICON plc, headquartered in Dublin, Ireland, it's all about data, mountains of it, that drives clinical trials, accelerates proof of efficacy, and hastens the delivery of lifesaving new medicines and therapeutic devices to market. As a global leader in contract clinical research, ICON's data-modeling team runs algorithms on a software-as-a-service grid platform deployed on NetApp E-Series systems. This advanced data modeling requires extreme performance.

AstraZeneca accelerates pharmaceutical science with the cloud. By partnering with NetApp, AstraZeneca has been able to design and implement a data strategy for its hybrid multicloud environment. This dynamic movement of data to any cloud from any cloud allows faster analysis and discovery.

NVIDIA Parabricks provides 30 to 50 times faster secondary analyses of sequencer-generated FASTQ files to variant call files. Additionally, Parabricks achieves results that are equivalent to the results of common secondary analysis tools like GATK4 and DeepVariant, while significantly increasing throughput. By using GPU-accelerated computing, Parabricks can provide throughput comparable to about 40 to 50 CPU servers with one GPU server, reducing IT management overhead and operating costs, including power and cooling.

In combination with NetApp ONTAP AI, Parabricks makes it possible to deploy a fully integrated, end-to-end AI solution tuned for compute and data-intensive genomics workloads. For example, in response to the impact of COVID-19, Core Scientific gave researchers free access to AI infrastructure as a service, powered by NetApp ONTAP AI and NVIDIA Parabricks, for GPU-accelerated coronavirus-related research.

These types of solutions are not prototypes. There are many real-life examples of AI improving genomics workloads—and helping patients. For example, San Diego researchers recently set a record for sequencing to diagnosis in a neonatal/pediatric ICU. The team created a pipelined workflow that included Illumina sequencers and Diploid Moon to automatically filter and rank the possible causative gene variants, compressing the entire workflow down to about 19 hours.

Many researchers are turning to the cloud to gain access to necessary storage and compute resources. Because of the large datasets and the amount of compute required, genetics researchers use a hybrid cloud approach to gain access to the necessary storage and other resources. In clinical settings, secondary and tertiary analysis of patient genomic data has to be done close to the sequencer, at least in urgent cases. Hospitals and laboratories that want to perform sequencing for clinical use will need to have local compute and high-performance storage, and the results of genetic analysis must be integrated with EHRs so that they are quickly available to clinicians.