The DNA Genome and Machine Learning

The DNA genome and machine learning would seem like an unlikely partnership since one is embedded in the biological world and the other in artificial intelligence. However, in recent times the human genome has been used as ‘training data’ for machine learning and has been able to predict the phenotype (in this case facial appearance) to a remarkable degree. (Photo from Riccardo Sabatini’s TED talk)


How was this done and how was machine learning used? Machine Learning (ML) is a "Field of study that gives computers the ability to learn without being explicitly programmed". My expertise in machine learning is somewhat limited and comes from a Coursera online course Machine Learning Foundations: A Case Study Approach from the University of Washington. My understanding of the process is illustrated in the diagram below (which may be only conceptually correct)

Machine Learning

The phenotypic characteristics of the subjects were codified and a random set was used along with their DNA sequences for ‘training’ the machine learning (ML) model.  A predictive model was made.  A set of ‘testing’ subjects was put through the model to evaluate the predictions. Further tweaking’ of the model was made through more iterations until some prediction endpoint was reached. This project involved the coordination of a large number of people working on different machine learning modules.

Watch the interesting TED talk by Riccardo Sabatini

At the end of Riccardo’s talk, he raised the issue that the human genome should be everyone's concern — philosophers, politicians, artists, scientists, business people and ordinary citizens. In the closing remarks of  my previous blog — I raised the concern: that since we have by and large eliminated ‘selection’ from the process of biological evolution,  we as a species shall continue to accumulate mutations in our genome — something that occurs on a daily basis. We potentially face an evolutionary dead end unless we are willing to intervene and correct these genetic mistakes. Also are we ready to grapple with the thorny problem of improving our genetic makeup?

One thought on “The DNA Genome and Machine Learning

  1. Pingback:

Leave a Reply

Your email address will not be published. Required fields are marked *