When Machine Learning and Biology Merge

Entering the Koch Institute at MIT earlier this December, I was pleased to be greeted yet again by the light bustling of students and researchers, the fast-closing elevator doors, and the faint smell of rat enclosures––though last was definitely not as pleasing.

My mission at MIT was simple––discuss the grounds for a new project involving AI for cancer research, specifically for the small intestine. Our team of five, including a computer science PhD, a cancer research professor at MIT, two undergraduate students, and myself, met to brainstorm.

We started out the meeting by simply asking ourselves just what we wanted out of this project. What could make research easier? What could make it more accurate? How can we spend less money? And, overall, how can we accomplish something that MIT has never seen before?

After getting a decent understanding of what we needed from AI at the Yilmaz lab, we began planning. Overall, we hypothesized that if we could feed AI multiple data sets, along with their context and conclusions, it could eventually make links between cancer triggers and potentially point out a new area of study that could yield discoveries. However, machine learning is a sphere that––though we've worked with and been exposed to it––none of us expertly know how to handle. We accepted that we may yield excellent results, mildly useful results, or fail entirely. We were at peace with that.

I certainly did not go into the meeting thinking that any of it would be easy. Asking ChatGPT to produce something is effortless enough––it's like waving a wand and having AI conjure exactly what you wish for. But working with deep neural networks, fine tuning an unpolished model, and––especially for me––gathering, organizing, and feeding the model the right data and the right amount of data––that is no quick feat.

This project is something that I will personally be dedicating a lot of my time to. It may work out or it may not, but I am simply grateful to be part of the process and excited to do a lot of new learning.

To help me with this journey, I started reading Introduction to Machine Learning with Python––A Guide for Data Scientists. It is an informational book written by Andreas C Muller and Sarah Guido that helps explain and teach machine learning to regular people that may have no previous experience with it. I will return with results if I find that it is helpful.