I'm Haron, a software developer, computer science student, and optimistic about Artificial Intelligence.
Before we get into details, let's get the unpleasantries out of the way:
- I don't have a degree in medicine or any medical related courses
- I was bad in chemistry and I never did biology
- I haven't done research about what I'm writing and all this is purely my thinking
I once had a crazy idea about how we could use Machine Learning to come up with cure for cancer - or a better way to manage cancer - and this idea can also be applied to diseases such as HIV/AIDS that don't have cure yet. But before that, a little history I know 😛
Based on my understanding so far, we have over 300 different strands of cancerous cells, and this makes the current treatment methods inefficient, because a patient can have different cancerous, and if a certain treatment method works for certain type of cancerous cells, it's not a guarantee it will work for all the cells. We've seen or heard a lot of cases, of people who were declared cancer free only for a few months or years down the line to be diagnosed with the same cancer they had before, and this can be attributed to the fact, that there are cancerous cells small enough to not get detected but they can persist for a long time.
Machine Learning has been used for early diagnosis of cancer. This helps because the earlier you diagnose the infection, the earlier the treatment the easier it it to control the infection. We can say this has drastically helped a lot of patients and saved a lot of lives, but still there are a lot of people who die every year because of the same. This only justifies the fact that we need to come up with new ways to manage this beast of a demon and it also justifies that my idea is not that crazy after all.
The idea aims to use artificial neural network (ANN), to come up with the cure for the disease. In a layman's language, we'll create a ANN model, train it with the relevant data, and finally use the model to predict the cure for cancer, depending on the models accuracy, we can make a compromise.
In details, the idea actually aims to collect the thought process of the virologist and rearchers who have come up with cure to these diseases in the past. The model would also be trained on the scienfic knowledge surrounding the production of medicine. In our basic structure, we humans are made up of chemical components, attoms that have a certain type of bonds. Our bodies react to chemicals differently. Based on the previous studies and patterns, our model can be able to map how different chemicals react to our bodies.
There is a reason why when we take a painkiller, it helps relieve pain, and this reason is not just biological, but chemical. There is a relationship between the painkiller and the effect it haves in our bodies. The same goes for all the medicine, and I'm sure it's safe to assume, this relationship were talking about can also be a pattern that a Machine Learning model can learn.
In it's basic form form, these medicine we take eg, pain killers, are made up of different chemical elements that react differently and it's their compound effect, that actually gives us the desired effect we want.
If we can create a model, with the thought process of researchers, but this models maps the different chemical components, with the different effect they have in our bodies, considering the side effects. If we also create a different model that will try to analyze the different cancer strands with the effect they have to our bodies. These models can be combined to predict a cure for cancer and other diseases without cure. Who knows?
Computers are really good in pattern matching, and they neved forget unlike us humans. With the right data, Machine Learning can be used to beat cancer. Imagine how many lives can be saved as a result. Thinking of that gives me the justification to chase this crazy idea 🤞