Machines learning promises us a future where hefty and complex data can be analyzed in no time and patterns can be logically drawn from an ocean of randomness. Apart from electronic medical records, the human gene and genome has also rendered a sizeable playing field.
Machine learning method is either supervised: training involving labeled DNA sequence which marks the start and end location of a gene; unsupervised: training which takes place without training data, or a hybrid of both.
The generalizable predictive nature of machine learning had enabled us to predict the impact of drug on a person with DNA mutation (i.e., pharmacogenomics), functional consequences of DNA mutations (i.e., pathogenicity prediction), how a locus mutation impact the expression level of a gene (eQTL mapping) and many more.
Finding defects in the gene before birth
The peak of this recent hype lies in Google’s DeepVariant release last November, when high-throughput sequencing (HTS) is now guaranteed with a greater accuracy via deep learning.
As machine learning in genomics gains popularity as quickly as the way direct to consumer genome sequencing package arrives at our door step, experts are questioning if babies should have their genomes sequenced. The collected data will form part of the person’s EHR.
Some parents may find the attempt pressuring as the screening for particular conditions like Down Syndrome became inevitable and they may be forced to end the pregnancy. On the other hand, sequencing the entire genome is likely to yield unknown data or insignificant result and generate needless anxiety.
If it’s not happening to mosquitos, it’s not happening to us
Interestingly, scientists are seldom satirical. Retrospectively thinking, if machine learning can help to generate medical meaning in a mess of gene and show us who will be the potential target of a form of cancer, perhaps the same logic can be applied to locate a termination gene?
British biotech firm Oxitec had engineered a genetic mutation into mosquito which makes them breed youths which are not able to survive, as a mean to deplete Zika virus. The way to do so is to release a male mosquito carrying the self-terminating gene and let it reproduce with female mosquito in natural setting.
However, because the innovation falls into the grey area of biotechnology and the ongoing debate if it acts as a pesticide or surely prevents the spread of Zika virus, the mosquito is not yet released.
Perhaps it’s just a matter of time before we discover a human terminating gene but before that, the question we need to address is, will lives be terminated unnecessarily if machine learning returns a positive result that a person bears the gene to a certain medical condition?