Machine Learning Helps Visualize DNA Repair Factors

A global consortium of researchers has used high-throughput microscopy and machine learning to identify new proteins involved in DNA repair, a methodology that could lay the foundation for new cancer drugs and improve the effectiveness of existing therapies.

Raul Mostoslavsky, MD, PhD, scientific co-director of the MGH Cancer Center and the Laurel Schwartz Professor of Oncology (Medicine) at Harvard Medical School, together with colleagues in Spain, US, Canada, and China, describe their highly sensitive method for visualizing DNA repair mechanisms in Cell Reports.

“We have in place exquisite mechanisms to repair DNA breaks, and when those fail, we end up with disease,” Mostoslavsky says. “We accumulate genomic instability, we accumulate mutations, and many diseases happen because of the inability of cells to repair DNA.” At the same time, drugs that interfere with the ability of DNA in cancer cells to repair themselves could halt replication and induce cell death.

Previous studies of DNA repair mechanisms used systems developed by biochemists to purify proteins, which had relatively low yields.  “We decided to develop a high-throughput assay to try to identify repair factors in a more unbiased way. We ended up developing a unique microscope-based automatic system to generate DNA damage and to collect information on proteins that are recruited to these types of damage.”

Combining this high-throughput microscopy with machine learning, the researchers were able to analyze how proteins are attracted to or excluded from double-strand DNA breaks. They then used the method to generate a library of 384 mostly unknown factors and single out which of these proteins are mobilized when DNA damage occurs. The team identified nine new proteins that are involved in DNA repair.

One application of the system could be improving the treatment of breast and ovarian cancers caused by mutations in the cancer susceptibility genes BRCA1 and BRCA2. These cancers are treated with a class of drugs known as PARP inhibitors that work by inhibiting a particular DNA repair factor.