Can AI Assess a Drug's Effectiveness?

Can AI Assess a Drug's Effectiveness?
Like

Share this post

Choose a social network to share with, or copy the shortened URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

I'm doing a project in which I use machine learning (considered a branch of Artificial Intelligence) to predict whether or not a drug may be effective at curing a disease.

Genes contain information to build proteins, molecules that perform a wide range of functions, including producing cells, defending the body, and communicating between cells. Diseases often target genes, which leads to the proteins being produced in a distorted way and leading to irregular functions. Drugs can come in and bind with the proteins, hopefully reversing the effects of the disease on the proteins. This is illustrated below:

Drugs (ligands) binding to proteins (receptors), like a "lock and key" or "hand and glove"
Drugs (ligands) binding with proteins like a "lock and key" or "hand in glove"

Lots of drug molecules are proposed during the early stages of drug discovery, but which ones would have strong potential to bind with the protein? This is where my project comes in. I'm attempting to create a machine learning model that would take the physical structures of the distorted protein and proposed drug molecule, and evaluate whether the drug molecule could bind with the distorted protein. This would allow scientists to narrow down what drug molecules to investigate further.

So far, the project has been challenging, but rewarding. It's an interesting task because of the unique intersection between technology and pharmacology (study of how drugs interact with the human body), and the challenges in understanding both fields. Earlier, I predominantly struggled with understanding the pharmacology side of the project: how do diseases, genes, proteins, and drugs link together? Later, it was more with the technology side of the project because I have constraints on how much computational space and power I can use, so I have to create a model complex enough to capture the intricacies of the problem, but not so complex that it takes a long time to train.

If anyone has any questions on what I am doing, they are more than welcome to contact me at ab550@st-andrews.ac.uk or ayambabu23@gmail.com!

Citation for image:

Joya, M.P. Protein-ligand binding, Protein-ligand Binding | BioRender Science Templates. Available at: https://www.biorender.com/template/protein-ligand-binding.

Please sign in

If you are a registered user on Laidlaw Scholars Network, please sign in

Go to the profile of Alexander Richardson
11 days ago

Great work Ayam!

Go to the profile of Ayam Babu
11 days ago

Thank you, Alex!