Google has created an AI program to generate drug discovery protein recognition. According to Venture Beat, members of Google Brain have crafted an artificial intelligence program capable of recognizing protein crystallization. The program already has accuracy rates estimated at 94%.
“Hundreds of experiments are typically run for each protein,” said Vincent Vanhoucke, a principal scientist at Google Brain. “And while the setup and imaging are mostly automated, finding individual protein crystals remains largely performed through visual inspection and thus prone to human error.”
Protein crystallization is what determines the shape of cells. It plays a critical role in the drug discovery, a multi-billion dollar industry, which is used to treat many different illnesses that affect both humans and animals.
For instance, there are more than 75 million pet dogs in the United States. Many dogs are affected by kidney disease and Lyme disease. Drug discovery is what allows chemists and scientists to look for cures for these illnesses.
“Critically, missing these [proteins] can result in lost opportunity for important biomedical discoveries for advancing the state of medicine,” Vanhoucke said.
Google researchers worked with the Machine Recognition of Crystallization Outcomes (MARCO) initiative to train the artificial intelligence model. MARCO is a partnership between academics and pharmaceutical companies.
This isn’t the first time Google researchers have shown an interest in improving healthcare.
In June 2018, Google researchers published their results regarding predictive deep learning models capable of determining the likelihood of a patient dying during their hospital stay, how long they’ll be in the hospital for, and how likely they are to be readmitted.
These learning models may be more accurate than traditional screening that’s often used for patients who are older and more susceptible to surgical complications.
Google also shared a computer vision model earlier this year that’s capable of detecting diabetic retinopathy, which is one of the top indicators of heart disease.
Many large tech companies and startups have been making their way into the healthcare industry. This is especially true in the imaging services where tech companies are more likely to thrive.
In fact, AI drug discovery startup Verge Genomics just recently raised $32 million in funding.
Baidu Research also announced this past June that it generated an algorithm capable of identifying tumor cells in breast tissue more accurately than pathologists. And although the algorithm has yet to be released, early results of the AI show that algorithm has been more accurate at detecting skin cancer.
The results of the work between MARCO and Google Brain are open-sourced and available on GitHub. The research is available in the journal PLOS One.
