2021 SEMINAR ARCHIVE
January 6th 2021
speakers: Mohammed AlQuraishi & Sergey Ovchinnikov
title: CASP14 summary
abstract: A group discussion about the results from CASP14 - similar to what we did two years ago for CASP13. We will unpack the dramatic results from AlphaFold2 and help us to understand their implications for the field of protein structure prediction.
February 3rd 2021
speaker: Chris Ing
title: The 7 Habits of Highly Successful Protein Therapeutics
abstract: A holistic, integrated, and principle-centered approach is necessary for any protein facing translational challenges in drug discovery. In the style of a self-help seminar for proteins, I reveal a step-by-step guide for becoming an effective and developable protein on the path to the clinic through penetrating insights and pointed anecdotes across academic and industry settings. Specifically, we'll cover the core design strategies (i.e. habits) necessary to adapt to change, attain therapeutic goals, and thrive as a protein therapeutic in the modern world.
March 3rd 2021
speaker: Ben Meinen
title: Rational de novo design of disulfide-rich miniprotein agonists and antagonists for family B G-protein coupled receptors
abstract: The family of B1 (secretin) GPCRs control physiological responses like glucose metabolism, cardiovascular and gastrointestinal development as well as immune response. Approaches to convert the intrinsically disordered endogenous peptide hormones into agonist and especially antagonist have been largely unsuccessful. Miniproteins are small 3-12 kDa proteins that have great promise for therapeutic applications, because they allow for the combination of the binding specificity of antibodies, the stability of small molecules, and a low molecular weight that improves tissue penetrance. We are using de novo protein design to engineer disulfide-rich miniproteins that act as agonist and antagonist for family B GPCRs.
April 4th 2021
speaker: Dylan Marshall
title: A view of the structure-fitness landscape from an unsupervised vantage
abstract: Inferring the structure and mutational fitness of a given protein from sequence patterns alone are two tasks commonly addressed with generative sequence models. Here, we pedantically highlight a dearth of consistency in mutational fitness assays and introduce pairwise-saliency, a novel method for revealing the structural information learned by a given model. We also compare and contrast the capacity of a suite of progressively complex models towards these two tasks, thus allowing a view of the structure-fitness landscape to come into focus.
speaker: Frances Chu
title: Designing c-KIT Receptor Inhibitors to Clear Hematopoietic Stem Cells Prior to Bone Marrow Transplants
abstract: Current methods to clear hematopoietic stem cells (HSCs) from the bone marrow prior to bone marrow transplants have adverse, off-target effects on patients. A preferable method is to inhibit the c-KIT receptor, a receptor tyrosine kinase responsible for HSC survival, proliferation, and differentiation. By creating a protocol to dissemble protein quaternary structure, an inhibitor of c-KIT is obtained from its natural ligand.
May 5th 2021
speaker: Jack Maguire
title: Autocomplete for Protein Design
abstract: Rosetta's "FastDesign" protein design meta-protocol is constructed as a linear series of rotamer substitution rounds. The field of machine learning has made strong developments recently in the area of natural language processing, which considers linear series of data. Let's merge these together! This project attempts to use introspective machine learning techniques to guide Rosetta design simulations to the finish line.
June 2nd 2021
speaker: Andrea Garavito
title: De novo design of mini-proteins to inhibit bacterial biofilm formation
abstract: Biofilms are a prominent health and economical threat to society because they can be found in almost every environment, including water supplies, air condition vents, oil pipelines, or even medical devices such as a prosthetic heart valve. The formation of the biofilms in many Gram-negative bacteria is mediated in the periplasm by the Lap system. We are using a novel computational approach to design de-novo mini-protein from a single hotspot residue to inhibit a protein-protein interaction critical to the Lap system. Here, I will discuss the computational approach and the experimental screening of the de-novo designs. This work will facilitate investigation of the mechanisms controlling biofilm formation, and will provide a demonstration that the Lap system is a viable target for biofilm inhibition.
July 7th 2021
speaker: Pallav Kosuri
title: New tools for studying the mechanics of single proteins
abstract: I will present my research into fundamental principles of protein disulfide bond formation, and tell you how disulfide chemistry may influence the mechanics of your muscles. I will then discuss my recent invention of DNA devices that can be used to amplify protein movements and visualize protein-DNA interactions. Finally I will share our future plans to develop mechanical toolkits for all sorts of molecules at my newly opened lab at the Salk Institute.
August 4th 2021
speaker: Sergey Ovchinnikov
title: ColabFold - Making Protein folding accessible to all via Google Colab!
abstract: We will walk through a series of notebooks designed to run RoseTTAFold and AlphaFold for protein structure and protein-protein-complex prediction.
September 1st 2021
speaker: Nazim Bouatta
title: Predicting protein structures from single sequences
abstract: Despite the outstanding performance of AlphaFold2 and RoseTTAFold in protein structure prediction, many challenges remain, including (i) prediction of orphan proteins for which multiple sequence alignments (MSAs) cannot be generated and (ii) understanding the physical principles encoding the rules of protein folding. In this talk, I will first provide a brief survey of the available deep learning methods for protein structure prediction and then describe our end-to-end differentiable model (RGN2) able to predict structures starting from single sequences and without MSAs.
October 6th 2021
speaker: Brian Coventry
title: Learning to design mini-proteins that bind to specific protein targets
abstract: Antibodies have no trouble binding specifically to nearly any protein, but how do they do it? We set out to discover the principle necessary to design mini-proteins that bind to arbitrary protein targets using only the target apo-crystal-structure as our guide. While we certainly can't match the broad spectrum binding of antibodies, for a certain class of protein targets (those with rigid, exposed hydrophobics), we've learned quite a bit and have found binders to 30+ targets so far.
November 3rd 2021
speaker: Anum Glasgow
title: How local changes in secondary structure flexibility can drive protein function
abstract: Transcription factors regulate genes in response to environmental signals. Evolution took 4 billion years to make hundreds of transcription factors that all have the same fold, but which respond to many different signals, posing the question of how ligand specificity is achieved. We determined the mechanism of specificity for one transcription factor using biophysical measurements and computational modeling. Learning how the conformational ensemble of the protein changes upon binding partners that drive different functional responses can help us to determine how it evolved and how we can re-engineer it for new genetic circuits.