#94 Killing Cancer with Machine Learning feat. Dr. Amit Deshwar
Aug 25, 2023
auto_awesome
Dr. Amit Deshwar, a scientist specializing in machine learning, discusses using machine learning to speed up drug discovery and fight diseases like cancer. They explore Amit's career path, his research on pan cancer analysis and RNA therapeutics, and the process of drug development using machine learning. They also touch on the advantages of being located in San Francisco for machine learning and share unique experiences and interests in the city.
Machine learning models are used in computational biology to analyze RNA sequences and predict outcomes.
Machine learning and whole-genome sequencing can help reconstruct the evolutionary history of tumors and identify therapeutic targets.
Machine learning is utilized to analyze genome-wide data and identify genetic variations associated with complex diseases, aiding in personalized treatments.
Deep dives
Using machine learning to understand the rules of RNA biology
In computational biology, machine learning models are being used to study the rules and patterns of RNA biology. By analyzing large datasets of RNA sequences, machine learning models can predict properties and outcomes based on DNA sequences. This allows for a better understanding of how RNA functions and how it contributes to genetic diseases.
Reconstructing the evolutionary history of tumors
Through the use of machine learning and whole-genome sequencing, researchers can reconstruct the evolutionary history of tumors. This involves analyzing thousands of tumor-normal pairs to determine the genetic changes that led to the development of a tumor. By understanding the order and timing of mutations, scientists can identify potential therapeutic targets and develop strategies for early cancer detection.
Applying machine learning to genetically complex diseases
Genetically complex diseases, such as diabetes and Alzheimer's, have unclear causality and involve multiple genetic factors. Machine learning models are used to analyze genome-wide data and identify genetic variations associated with these diseases. This knowledge helps in developing genetically targeted therapies and personalized treatments.
Utilizing transfer learning for improved predictions
Transfer learning plays a crucial role in computational biology. By training models on large datasets from related problems (e.g., studying drug effects on cells in a dish), researchers can transfer the learned rules to make predictions in more complex systems (e.g., cells in the human body). This approach combines available data sources and improves the accuracy and applicability of predictions.
The Importance of Data in Computational Biology
Data is crucial in computational biology, especially when applying machine learning models. However, accessing high-quality data sets in the right cell types and environments is a challenge. While some publicly available data sets like the UK Biobank and GTEX exist, researchers often have to generate their own data sets to build the desired models.
RNA Therapeutics and Steric Blocking Oligonucleotides
RNA therapeutics, such as mRNA vaccines, have been successful in delivering specific instructions to produce desired proteins in the body. In contrast, the company mentioned in the podcast uses steric blocking oligonucleotides, which bind to naturally occurring RNA and prevent the binding of other proteins or RNAs. These oligonucleotides can increase gene expression in a targeted manner and are chemically synthesized using automated techniques. Although still in the pre-clinical stage, this approach has the potential to revolutionize drug delivery and treat various genetic diseases.
#94 Killing Cancer with Machine Learning with Dr. Amit Deshwar
Today I'm joined by Dr. Amit Deshwar. He uses machine learning to discover new drugs to cure various diseases including cancer. He's a scientist who works in the growing field of Computational Biology, and has risen through the ranks at the Canadian biotech company Deep Genomics.
During College, Amit got two internships at Google as a platform engineer. He then decided rather than working in big tech he wanted to go back to school and get his Ph.D. He studied Electrical Engineering and Computer Engineering at the University of Toronto, and had his work published in Nature, one of the most prestigious scientific journals.
I met up with Amit at the Glen Park library in San Francisco, at the exact table where the FBI arrested notorious Slik Road Darknet marketplace founder Ross Ulbricht.
We talk about how scientists and developers use machine learning to speed up drug discovery. I ask him a lot of my totally naive questions about how these therapies work and how they can fight various types of cancer and other diseases.
Photo of Amit arresating me at the Glen Park Library where the FBI arrested Ross Ulbright: https://drive.google.com/file/d/15B8HD4SGErnOd8zA-9gYW2MabAQFG58Q/view?usp=sharing
Photo of me arresting Amit: https://drive.google.com/file/d/1OWyaVyzqT8YgLFYUVi5kqY9te6ShSdgr/view?usp=sharing
Amit on Google Scholar: https://scholar.google.com/citations?hl=en&user=QGCYxysAAAAJ
