My project – in progress – is using machine learning methods to use clinical literature to classify a genetic mutation as a driver or a passenger of tumors. This innovation is important because clinical pathologists currently have to identify which mutations contribute to the growth of malignant tumors and which are neutral and have no effect. They have to review and classify each based on clinical literature which is incredibly time consuming. This is personalized medicine as it is different for every individual which makes reading all of this literature necessary to help the patient is almost an impossible task. The large scale effect of this would be that more lives will be saved that would have been lost to cancer. By identifying the genetic variants, one would be able to earlier detect firstly – if someone has a certain cancer gene, and then secondly – identify it early enough that it will be treatable. Determining the malignancy of a genetic mutation is key to providing personalized care to patients. My project will work in the following way: the model will give each clinical text a classification: a number between 1- 9 which represent different statuses of the mutation. For example, 1 means that it has a likely loss of function, 2 means a likely gain of function, and 3 means neutral… My model uses python code, but may mainly built on tensor flow when I get to that stage. This is a small step, leading to a better future.
What inspired you (or your team)?
My inspiration on this project was influenced by many things. First of all, it combines my interest in both computer programming – mainly the mathematical aspects, and medicine. I used to do lots of personal exploration projects in elementary and middle school as well, mainly targeted at stem cells and basic genetics. When I did these projects I ended up reading lots of journal articles, and some of them really spoke to me. I would like to be able to help people – and this is the area in which I plan to do it. If they had I think I would enjoy working in bioinformatics in the future, so I figured this would be a good insight into whether I would enjoy that field.