Найти в Дзене
3 подписчика

Artificial Intelligence Will Help People with Pharmaceuticals


When scientists try to diagnose and treat a disease, they often look for mutations in a single gene that causes the problem. Or they study the average effects of a mutation that cause a disease across a generation. However, both of these approaches ignore the complexity and specificity of other factors that cause disease to spread — demographic information, proteins, multiple gene interactions, environmental exposures, and more.

Until recently, computers weren’t powerful enough to analyze all of this health information; there weren’t enough data sets to analyze either. But advances in artificial intelligence could spur interactions with large medical data sets, including the ability to quickly sequence entire genomes and quickly mine molecular information. AI could make precision medicine a reality, and one day it could learn to identify the unique characteristics of an individual that may lead to certain diseases and determine how to treat them.

'That’s what precision medicine is all about. Each of us is different, each of us is unique genetically, so each of us should be treated in a way that takes into account our genetic diversity and our environmental history,' says Jason Moore, head of computer science at the University of Pennsylvania. 'So I think where AI can be useful is when it can bring together multiple genetic and environmental factors to identify subgroups that are important.' Two scientists, including Moore, presented their AI work at a conference on Big Data and Predictive Knowledge for Disease Control at the New York Academy of Sciences. Medical AI will allow computers to 'think' about genomics, disease, and treatment like humans do, only faster, better, and on a larger scale.

One of the most exciting applications of AI is refining the actions of new drugs, something that previous methods have lacked. With the average drug development taking up to 14 years and costing around $2.6 billion (at least overseas), pharmaceutical companies are willing to do anything to reduce that time and cost.

Dr. Niven Narain, co-founder, president, and chief technology officer of biopharmaceutical company Berg, described the company’s Interrogative Biology AI platform, which identified several drug targets that have been in development for more than 25 years. Berg’s platform collected as much information as possible about patients — from demographics and environmental conditions to genetic mutations — to identify opportunities for new treatments. Narain says Berg’s method has cut the time and money needed to develop drugs by more than half.

'Not only are we reducing the time it takes to make a drug, but the drug we make will have an enhanced effect,' Narain says. 'That’s hard to overstate because if you make a drug, let’s say it helps 10,000 people. But if you make it faster, even with AI, it could help 10,000,000 people, which is a big difference.'

Using its AI system, called EMERGENT, Moore’s lab found five new biomarkers that could be potential drug targets for glaucoma. To do so, the company needed input from 2,300 healthy and diseased individuals, information on 600,000 individual DNA sequences, and knowledge of specific gene interactions; they fed all of that into EMERGENT. One of the DNA sequences identified by the AI system caused glaucoma; five others provided new drug development opportunities.

Next, Moore says, his group is working on better ways to visualize the data that AI computers can produce — the results won’t be useful unless biologists can interpret them in any way they see fit. Most interestingly, his group is using the Unity 3D video game platform to develop applications that allow scientists to fully immerse themselves in data and AI algorithms within the game system.

'Imagine if all your big data lives in a video game, and you’re flying through it and you find something interesting. Within the visualization, you want to say, ‘Aha, this looks interesting,’ and click a button to run an analysis on a piece of data you’ve see
yt3.ggph
Artificial Intelligence Will Help People with Pharmaceuticals When scientists try to diagnose and treat a disease, they often look for mutations in a single gene that causes the problem.
3 минуты