Because the brain’s inner workings remain impervious to deep understanding, addressing psychiatric disorders presents a particular challenge. But a provisional clarity may be emerging. Neuroscientists and computational biologists are making headway in determining the genetic roots of illnesses like schizophrenia, which could lead to more targeted treatments.
The ongoing PsychENCODE project, which operates at several medical centers and is funded by the National Institute of Mental Health, is one such undertaking. Working from a sample of nearly 2,000 brains from deceased individuals—roughly half of whom had no diagnosed psychiatric disorder, and half of whom did—the PsychENCODE team focuses on “noncoding” DNA, which makes up 99 percent of our DNA. Some noncoding DNA tells our genes how to operate, including the genes that will manifest as schizophrenia in some people. PsychENCODE researchers want to know exactly which variants of noncoding DNA drive this condition.
In December 2018, PsychENCODE researchers published a collection of papers describing their progress. One paper, coauthored by Mark Gerstein, a professor of molecular biophysics and biochemistry at Yale, describes the development of a “deep-learning model” that predicts which variants in noncoding DNA are more likely to interact with people’s genes in such a way that schizophrenia sometimes results. Although this is not a bull’s-eye prediction of who will become schizophrenic and who will not, Gerstein argues that the deep-learning model has better predictive power than other methods.
The model, a form of artificial intelligence, aims to use abstract knowledge gained in the research lab to improve clinical treatments for real patients. The ultimate goal, says Gerstein, is to use the model to develop pharmaceutical treatments that reduce the impact of schizophrenia. Part of the challenge in developing drugs to treat the disease is the fact that it is not a one-size-fits-all condition.
Permission required for reprinting, reproducing, or other uses.