Externally contributed article
Author: Artur Meyster
Machine learning, a branch of artificial intelligence, allows data scientists or engineers to design algorithms to train computers and machines – in a sense, to teach them how to learn.
It has many applications in many niches and is transforming genetics. Machine learning on its own is a reasonably new technology, but it has been able to quickly reach new limits thanks to the latest technological developments in computer power.
Let’s explore the latest applications of machine learning in genetics.
In the pharmaceutical industry, pharmacy genomics, or pharmacogenomics, is the study of the genomes response to drugs. In other words, it explores how a person responds to certain drugs, depending on their genetics.
Scientists have already completed research into applying machine learning techniques to pharmacy genomics, like artificial neural networks or regression trees, to determine personalized doses for new drugs.
IBM is also combining genomic tumor sequencing and cognitive computing to give cancer patients personalized treatment. The software is called Watson for Genomics and it digests more than 10,000 medical articles and 100 clinical trials per month. The algorithms compare the genomic tumor analysis to all the databases to recommend the best treatment possible.
Gene editing is the ability to modify the genetic information of any living organism, such as plants, animals or even humans. The process consists of replacing one DNA sequence with another. DNA sequencing used to take a lot of time; each genome has a lot of data that needs to be analyzed. Now with machine learning techniques, sequencing DNA can be done in just one day.
Since 2012, with the discovery of CRISPR-Cas9, scientists have been able to modify DNA to either kill cells, like cancer cells, or modify genetic errors like Down Syndrome. Scientists first have to know which DNA sequence they want to change before editing the genes. That’s why combining machine learning with CRISPR-Cas9 improves efficiency.
As mentioned, plants can be genetically modified. It is the most common application of gene editing in the industry. For some years now, the agriculture industry has been genetically modifying crops to increase production levels and physical appearances like size and color.
Machine learning can also be used to improve soil quality and analyze genetic information to prevent and predict diseases. For example, the Australian organization CSIRO conducted a study where they integrated genome information with field performance data and climate records using machine learning. This provided Australia’s agricultural industry with user-friendly insights about specific crop performance.
Consumer Genomic Products
The ancestry test has become not only a trend but nearly ubiquitous. Dozens of companies are taking advantage of the technology to trace family lineage as demand for the DNA tests continues to rise.
To take part, clients send a sample of their DNA and receive a complete breakdown of their DNA origins and where their ancestors were from. Although the tests aren’t 100 percent accurate, they get closer to the truth with every new machine learning development.
Many of the companies offering the tests have been founded in the last few years and use unique algorithms to make predictions as accurate as possible. The process for these tests is straightforward – people buy a test kit and send their DNA sample to the lab. Then the company discovers the unique traits of the DNA while comparing it with their database.
Each company’s database grows as more and more customers take part in the tests. This way, their algorithms can make accurate predictions of which region in the world your DNA came from. The bigger the database, the more accurate the predictions. Using machine learning applications, the potential for complete accuracy rises as the companies expand.
About the author
Artur Meyster is the CTO of Career Karma (YC W19), an online marketplace that matches career switchers with coding bootcamps.
He is also the host of the Breaking Into Startups podcast, which features people with non-traditional backgrounds who broke into tech.