In the eye of the Argonne National Laboratory, a multidisciplinary science and engineering research institution in Illinois, precision oncology medicine shares something in common with longer-lasting batteries and origin and evolution of the universe: they are challenging topics that many researchers are willing to devote their lives on. Indeed, according to a MIT Technology Review article published early this year, there are “potentially more drug-like molecules than the number of atoms in the solar system”.
As such, finding a drug which cancer patients best respond to, is testing the limits of modern science, until, probably the emergence of artificial intelligence (AI). Some scientists couple machine learning and genomics to sequence data and help clinicians to better understand how best to tailor treatment plans to individual patients. Others leverage on cognitive computing: requiring computer to mimic the way human brain works, to make sense of biomedical datasets and uncover novel oncology drugs.
A new AI chip
Last week, Argonne announced its official deployment of a new AI processor, CS-1, developed by computer systems startup Cerebras, to increase the rate of training deep learning algorithms. CS-1 houses the fastest and largest AI chip ever built and it is regarded as the new generation of hardware that will accelerate AI development.
As Rick Stevens, Argonne Associate Laboratory Director for Computing, Environment, and Life Sciences said in the press release, “by deploying the CS-1, we have dramatically shrunk training time across neural networks, allowing our researchers to be vastly more productive to make strong advances across deep learning research in cancer, traumatic brain injury and many other areas important to society today and in the years to come”.
At the moment, graphical processing units (GPUs) are the most common chips used in deep learning. GPUs were popularized in games and graphic productions because of their abilities to speedily generate pixels. That is also the reason why they were selected to be adopted by the AI world in the first place. However, GPUs are not meant for specialized AI training and they consume a large amount of energy in the process of training deep learning algorithms.
As such, companies ranging from Intel, Nvidia, to startups like SambaNova, Groq and Graphcore, are fiercely competing with one another to come up with a new chip that is well-catered to the growing demands of AI development.
Cancer drug research
CS-1 was chosen by Argonne not only because it is a lot faster than the present general purpose processors but also because it has the capabilities to handle scientific data, in a reliable yet easy to use manner. “We have a lot of higher-dimensional data sets” Stevens told MIT Technology Review. These data usually come from rather diverse data sources and the developed deep learning algorithms are extremely complex as compared to computer vision or language applications.
Thus, the main duties of CS-1 is to dramatically increase the speed of developing and deploying new cancer drug models. Argonne is hoping CS-1 will come up with a deep learning model that can predict how a tumor may response to a drug or a combinations of two or more drugs. Moreover, Argonne also wishes the new model is capable of generate new drug candidate which produces intended effects on a particular type of tumor and predict the effects of a single drug candidate on multiple types of tumors.
A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.