Over two decades ago, in 2003, the Human Genome Project reached its goal, and developed the full human genome sequence. And though one milestone was reached, genomics has not stagnated. Now, with the development of machine learning, and particularly with the recent breakthroughs and associated AI “fever,” we have an opportunity to look at the many roles that AI might play within the field of genomics (as long as none involve a robot revolution).
Already, many technologies, both developed and in progress, connect hidden patterns within genomic data to physical information. These technologies have many applications — from using facial analysis to identify genetic disorders, to distinguishing pathogenic genomic mutations from the benign ones. Deep learning can be used to improve CRISPR tools; it has helped in analyzing the most effective disease treatment based on genomic data.
Each new application brings with it advances in the technology itself. Evaluation of AI used to identify Type 1 diabetes revealed that genotypic data is most useful in classifying diseases which most frequently result from single nucleotide polymorphisms. Research into Crohn’s disease found that non-linear classification methods are more successful at identifying disease. These steps bring the field closer to widespread availability, and to new applications of machine learning within other fields of genomics and medical diagnosis.
Much of the current ML technology lies within cancer genomics. In 2019, researchers developed a model to diagnose cancer type non-invasively. Using blood samples, their ML model ranged from a 57% to 99% success rate; in combination with another non-invasive technique, they successfully found 91% of patients with cancer. Another innovation analyzes the progression of cancer from tumor sequences. This technology could aid in predicting disease progression before it occurs – potentially creating new treatment opportunities.
AI has the greatest potential within the fast-growing field of personalized medicine. Using genomic data, medicine can be specialized to the needs and sensitivities of the patient, creating more efficient and effective solutions. Some patients have already begun to receive genomics-based care, and this sci-fi -like idea is approaching reality. Through machine learning, genetic data can become a path to medical diagnoses and treatments.
Comments