HTGAA IX: Protein Design (online)
In the recent years, our ability to engineer biomolecules with novel function has undergone a rapid change. This is due to an influx of computational tools and models for predicting protein function and structure. However designing proteins that function in the lab as intended is still a challenge, and takes multiple rounds of iteration.
The field of protein engineering is incredibly vast, encompassing proteins with many diverse functions, many of which lack easily deployable assays. This is crucial to consider when developing a pipeline, as it can influence whether a computational approach is suitable and what types of in-silico metrics to consider while selecting candidate sequences for experimental verification.
In this recitation,
We will explore machine learning tools for protein design.
How to build a pipeline using these tools that works synergistically with the assay being used to test for function.
Balancing Compute Time vs Experimental Costs
Case studies and examples of applying these tools for protein engineering.
Ideas for protein engineering projects and where to start
There are multiple models that do the same deep learning task like protein structure prediction, sequence recovery etc. Here we have a living document of current models and tools and most of them have easy to use colab notebooks. To answer the questions in Part C choose any of the models. You can also use all of them if you want to evaluate their performance across various tasks just note that these models take a long time to run on colab.
Meeting link to be sent upon registration. A big thanks to Manu Ponnapati for helping teach this class!