By Misbah Uddin
An organism’s genetic code could be the difference between advancing the species, to killing it off in the long run. As animals big or small grow and reproduce, their DNA mixes and spreads to their offspring, creating similar yet distinct gene codes for every individual. But the process is never perfect, so where one individual may produce a protein necessary to survive, another could be born with no ability to make that same protein and will die off, even due to just 1 mutation in their gene code.
For nearly a century, scientists have tried to map out these mutations, trying to predict how a gene code would mutate based on selection pressure, and what kind of gene expression would occur on these “fitness maps”. It was a limited concept however, until recently when researchers successfully created an AI Neural Network model to make these predictions for them, making a visual map that could detect signatures of genetic change and gene regulation.
MIT graduate researchers had fed their AI Network with hundreds of millions of random but known 80-base pair sequences in yeast, each sequence coding for a yellow fluorescent light with varying levels of expression. As the AI analyzed each sequence, it slowly recognized patterns the human eye could not, allowing it to predict how the genome could change with time and how the aforementioned trait could be affected by those changes. The AI was also given real life population data sets for yeast as well, only further specifying the gene regulation calculations the network was capable of modeling.
With all this data, researchers were able to create an “oracle” for scientists to reference back to for both the history of genetic changes, and it’\s use in the future. The population data that the AI received unlocked thousands of years of selective pressure on yeast, allowing us to understand what pressures sculpted these gene changes. This understanding is also useful in discovering diseases, knowing what specific codes would cause and therefore treat certain afflictions and illnesses. Companies producing enzymes, antibiotics, and in food production could also use the network to know how to manipulate genomes in order to express the best level of desired traits.
Comments