Google’s AI technology will aid researchers in understanding how genes function

Google’s AI technology will aid researchers in understanding how genes function
  • Homepage
  • >
  • All News
  • >
  • Google’s AI technology will aid researchers in understanding how genes function

Scientists sequenced the first human genome in 2003. They revealed all the DNA instructions needed to make an individual. We still did not know the function of all 3 billion letters in our genome.

Google DeepMind says that it has made an important step in its quest to crack the code. AlphaGenome is an AI model which predicts how small DNA changes will affect a variety of molecular functions, including whether or not a gene will become more active. This is the type of question that biologists are constantly assessing in laboratory experiments.

Pushmeet Khouri, vice president of research at DeepMind, says, “We’ve created for the first-time a model that unifies the many challenges associated with the understanding of the genome.”

AlphaFold is a 3D protein shape prediction technology that was released by Google AI five years ago. This work won a Nobel Prize in 2013. It spawned Isomorphic Labs and many companies hoping AI can find new drugs.

AlphaGenome is a project that aims to simplify biologists’ work. It does this by answering questions regarding how DNA changes affect gene activity, and how, ultimately, genetic mutations can impact our health.

Caleb Lareau is a computational scientist at Memorial Sloan Kettering Cancer Center, who had access to AlphaGenome before anyone else. This is by far the most advanced tool available to simulate that.

Google has said that AlphaGenome is free to non-commercial users, and it plans to reveal the full model details in the near future. Kohli said that the company was exploring how to allow commercial entities, such as biotech firms, to use this model.

AlphaGenome, according to Lareau, will enable certain experiments that are currently done in a lab to be performed virtually on a computer. Studies of DNA donors often reveal thousands of differences in genetics, which can affect the likelihood of a patient developing a particular disease, such as Alzheimer’s.

DeepMind software, says Lareau, could quickly predict how these variants work at the molecular scale. This would have otherwise required time-consuming laboratory experiments. He says that he will get a list of genes variants and then tries to determine which ones are doing anything. This system helps us to make a first-guess about the effects of any gene variant when observed in humans.

AlphaGenome won’t predict much about an individual, but it will give you a lot of information. AlphaGenome provides clues about the molecular activity of genes, but not like 23andMe which reveals a person’s traits and ancestry.

Google stated that “We’ve not designed or validated AlphaGenome to predict personal genomes, which is a challenge known for AI models.”

The AI system relies on the Google-invented “transformer architecture” that powers other large language models, such as GPT-4. The data used to train this model was a vast collection of public science projects’ experimental data.

Lareau claims that the new system won’t change his daily lab work, but it could allow for new research types. Sometimes doctors come across patients who have ultra-rare forms of cancer with a plethora of unfamiliar mutations. AlphaGenome can tell you which mutations cause the problem and possibly point to a solution.

According to Julien Gagneur of the Technical University of Munich, “a hallmark of cancer” is when specific DNA mutations cause the wrong genes to express themselves in the wrong circumstances. This type of tool can be used to narrow down the genes that are causing improper gene expression.

Patients with rare genetic diseases, who may never know the cause of their disease, could also benefit from this approach, even though their DNA is decoded. Gagneur says that while we can get their genomes, they are unable to tell us which mutations cause disease. AlphaGenome, he believes, could provide medical researchers with a better way to diagnose these cases.

Some researchers hope to eventually use AI in order to create entire genomes and new forms of life. Some researchers believe that the models can be used in a virtual lab for drug testing. Demis Hassabis said in this year that he would like to create a “virtual cell”.

Kohli describes AlphaGenome as a “milestone on the way to this kind of system”. He says that AlphaGenome does not represent the entire cell, but is beginning to shed some light on DNA’s broader semantics.

View Article Source

Share Article
Facebook
LinkedIn
X
Berkshire heating engineers put the boiler service first as January bill forecasts climb
As Christmas lists get shorter, Ascot home brand says one lasting piece beats five impulse buys
Ofcom names four age-check methods to underpin the under-16 social media ban
Secret Link