NVIDIA to Create First AI Framework to Assist Cancer Research

NVIDIA will team up with the National Cancer Institute, the U.S. Department of Energy, and several laboratories to make an AI (artificial intelligence) system to boost cancer research efforts.

This is part of an initiative called the Cancer Monshoot, which was announced by President Barack Obama during the 2016 State of the Union Address. It aims to deliver a decade's worth of advances in the field of cancer prevention, diagnosis, and treatment. The effort will focus on building a system called the CANDLE (Cancer Distributed Learning Environment) which will help researchers fight against cancer.

According to Yahoo!, the system will be the first framework designed to battle cancer with the help of analysis. CANDLE should be powerful enough to perform complex tasks that researchers would find difficult to do.

Various engineers and computational scientists will be working on developing the software that will be optimized for the latest computing infrastructure. The goal is to make CANDLE help generate 10-times annual increases in the productivity for cancer researchers.

Rick Stevens of the Argonne National Laboratory said that this is possible because new computing architectures have already accelerated the training of neural networks by 50-times in the past three years.

Jen-Hsun Huang of NVIDIA added GPU deep learning has provided analysts with tools to tackle various "grand challenges" that have been proven difficult for even supercomputers.

The Cancer Monshoot partnership will accelerate precision oncology projects. This will help provide a better understanding of cancer growth, discover less toxic therapies, and understand key drivers to effectiveness outside clinical trials.

CANDLE will do this by discovering the underlying genetic signatures in DNA and RNA of common cancers and relate it to the data collected by the NCI. It will then accelerate the molecular dynamic simulations of key protein interactions to see the underlying mechanics creating cancer conditions. Lastly, CANDLE will utilize semi-supervised learning to automate information extraction and analysis of millions of records to build a surveillance database.