Description:

It takes over a decade for a drug to reach the market. The cost of Phase 2 and Phase 3 trials has exponentially increased in millions since early 2000s. This increase in costs and production time has reduced the willingness of drug companies for clinical trials. A pharma company has to bear the undetermined costs and its researchers strain to evaluate therapeutic effects of novel drugs. A roadmap to a fast and affordable production of efficacious drugs is the solution. Advancements in ML has allowed medicinal chemists to generate novel molecules. It has allowed them to design it to their needs in a short time. I would like to present a paper on the current use of DL in drug discovery, the molecular representation of drugs for computation, the models used, the challenges it faces, the opportunities we have and some examples. It would brief about the pipeline for this process.

Prerequisites:

The audience need a layman understanding of how generative models work and just assume molecules need certain properties to function efficiently. I have written a blog on it in my medium account which you can check it out. I have referenced the paper and some blogs that made me grasp the concept in it . Just be curious and use ChatGPT, when you feel confused.

Content URLs:

https://medium.com/@pvikraman86/generative-molecular-design-7a8f83afbc59

https://www.nature.com/articles/s42256-024-00843-5

https://lilianweng.github.io/posts/2018-10-13-flow-models/

https://lilianweng.github.io/posts/2018-08-12-vae/

Speaker Info:

I am a 4th year CSE undergrad at SRMIST. I am passionate about many topics but currently DL for chemistry sparks my interest. I have did a research internship on Knowledge Graphs from IISER Bhopal and attended a summer school on ML for Healthcare at ETH, Zurich.

Speaker Links:

https://www.linkedin.com/in/vikraman-p-332549176/

https://medium.com/@pvikraman86

https://github.com/infinity-void6