Rosetta Commons Research Experience for Undergraduates (REU)

"Design of protein binders to inhibit and stabilize protein interactions to disrupt viral lifecycles"

Viruses, such as Human Immunodeficiency Virus (HIV), are highly disruptive to society in terms of health care costs and lives. The lifecycle of HIV involves the dynamic packaging and unpackaging of its genome into a capsid of proteins. Over stabilizing the assembly of the capsid proteins is hypothesized to disrupt the proper assembly or disassembly of the capsid/matrix and therefore inhibit viral lifecycle. We propose to utilize deep learning protein design methods to develop high affinity binders targeting an assembly of capsid proteins thereby over stabilizing them.

“Antibody engineering with deep learning”

Antibodies are critical molecules in the immune system and as pharmaceuticals. But many deep learning methods perform weakly on them because of their great diversity and reliance on loop structures, which have fewer data available. In this project you will create or apply new approaches such as language models, diffusion or flow models, toward predicting and designing better antibodies.

"Protein engineering with deep learning"

In most hub proteins, such as P53, the region that binds to different proteins lacks a 3D shape, but gets its shape upon binding to different targets: S100B, CPB, SIRT1. When it is cancerous, its affinity changes to different protein targets, depending on which protein it is binding to. We are trying to design inhibitors against targets rather than the P53 itself to study the role of each one of them alone.

"Protein design using generative models"

Combinations of novel AI design methods and the generative nature of RFdiffusion are converting previously intractable protein modeling and design problems into tasks that scientists can solve within weeks. Students will learn cutting edge deep learning protein design methods, and apply them to current design challenges. Areas of focus will include de novo enzyme design and de novo binder design. By the end of the experience, students will understand the computational pipeline of binder design and protein engineering experimental techniques. They will develop proficiency in molecular modeling, machine learning, and protein structure prediction techniques.

Alternative dates for this location: June 1- June 5 for training and June 21 - August 22

"Deep learning guided design of targeted protein editors "

There are more than 20,000 different proteins in or on a human cell. The ability to precisely target and enzymatically modify these proteins in programmed ways could allow unprecedented ability to interrogate and intervene in their biology. Advances in deep learning have enabled the design of binding proteins, particularly to targets with a well defined three dimensional structure. Building on these advances, we are developing methods to design affinity clamps that selectively bind unstructured C-terminal tails of proteins using engineered versions of natural tail binding proteins called PDZ domains. We are also designing protease enzymes with tailored specificity to modify any target protein at a chosen location. These bespoke PDZ-Protease molecules are expected to deliver a "one-two punch" to any target protein of choice. This design blueprint of our protein editors is reminescent of CRISPR technology for genome editing.

"Expanding the functional space of de novo designed protein assemblies"

Symmetry is widely explored by biological systems, enabling the formation of large complexes from multiple copies of the same protein building block, thus expanding the functional capabilities beyond those of the monomer itself. My lab focuses on the development of protein design methods to expand the structural and functional repertoire of protein assemblies available for synthetic biology and biomedical applications. Our goal is to understand the biophysical rules guiding self-assembly and leverage them to create bioactive, responsive protein-based materials with unprecedented programmability.

" Rigid feet for walkers"

Proteins are nature's nano-machines, driving essential biological processes, including those involving mechanical work: DNA replication, membrane remodelling, and cargo transport. Synthetic protein motors have enormous potential in health and material applications and could help us unravel the mechanism of natural protein motors.
We will combine cutting-edge AI design tools (RFDiffusion, ProteinMPPN, AlphaFold2) with single molecule tracking experiments to produce a completely de novo designed protein walker system with rigid feet. The incorporation of rigid feet enables an asymmetric interaction energy landscape that is a necessary precondition for powered walking. Powered molecular robots will open new frontiers in programmable materials and targeted therapeutic delivery.

“Elucidation of esterase mechanisms”

The intern will use various modeling tools to design esterases with improved stability to process-like conditions, then express and characterize their activity (mechanism, kinetics, selectivity).

“Predictive modeling and design of cyclic peptides”

In antibody drug discovery, two important goals are to improve antigen binding while reducing antibody self-interactions, and modeling is useful in prioritizing engineering efforts. We have generated large datasets along with homology models and conformational ensembles for each antibody in the dataset. The successful student will leverage Rosetta to generate structure-based descriptors and use them in building predictive machine learning models. The student will work with Merck & Co. scientists to assess the advantages of derived predictions alone and in combination with state-of-the-art predictive approaches.

“Design and engineering of Therapeutic miniproteins”

Miniproteins represent an emerging interesting modality with potential in therapeutic applications. The student will use computational design methods to engineer one or more attributes that are important in therapeutic molecules; affinity, selectivity, stability, solubility...etc. The student will be embedded in the biologics discovery department in a pharmaceutical company and will additionally learn the general processes involved in the discovery of human therapeutics.

"Physics-based and machine-learning methods to generate ensembles of membrane protein backbones"

The oligomerization of protein macromolecules on cell membranes plays a fundamental role in regulating cellular function. From modulating signal transduction to directing immune response, membrane proteins (MPs) play a crucial role in biological processes and are often the target of many pharmaceutical drugs. Despite their biological relevance, the challenges in experimental determination have hampered the structural availability of membrane proteins and their complexes. Computational docking offers a promising alternative for modeling membrane protein complex structures. The goal of this project is to compare computational tools for capturing the flexibility of membrane protein backbone, which often hinders accurate protein-protein interface predictions.

"Design of cellular therapies through computational protein design"

Chimeric antigen receptors (CARs) are the major antigen recognition domain for CAR T cells to detect cancer cells in the tumor microenvironment. The design of the CARs’ affinity and its biophysical properties allow for the fine-tuning of the CAR T cells effector function. Here, we will employ a combination of AI tools and classic biophysical tools to probe and manipulate CAR properties to design the desired affinity and effector function of the CAR T cell. The project can be both conducted as a purely computational project as also as a combined computational and wetlab project.

"Protein design using ML methods, high throughput screening for protein properties"

Our lab pursues a range of interdisciplinary projects at the interface of structural biology, immunology, and protein engineering. Current efforts focus on the design of protein-based vaccines, antiviral therapeutics, and molecular machines for gene delivery, as well as the development of immune modulators. These projects integrate computational modeling, protein design, and experimental validation, offering a diverse training environment. Depending on the intern's interests and the current status of ongoing projects, any of these focus areas could form the basis for an REU project. Interns will have the opportunity to contribute to cutting-edge research while learning methods in protein design and engineering, virology, and immunoassay development.

"Improving Design and Selection of De Novo Therapeutic Peptides"

Peptide therapeutics are emerging as one of the most promising new drug modalities. This project focuses on improving computational methods for de novo peptide drug design, integrating machine learning, Rosetta modeling, and other physics-based approaches to identify and evaluate novel protein–peptide interfaces. The intern will gain experience in the drug discovery industry, collaborating with a cross-functional team that includes specialists in AI/ML, biophysics, and experimental peptide synthesis and assays. Depending on project needs and interests, potential directions may include designing and evaluating novel peptide candidates or developing tools and workflows that enhance the efficiency and quality of the design process.

"Cracking the human immune repertoire using AI"

We are a (largely) experimental lab with an audacious goal of predicting and designing how antibodies can interact with new molecular surfaces. AI protocols are poorly suited for this problem because antibodies are sufficiently distinct enough from the rest of the proteome. My group is developing and implementing new assays to provide the necessary experimental data for deep learning methods to capture antibody molecular recognition.

"Design of ion channel modulators"

The overreaching goals of my research are: 1) to rationally design novel subtype-specific ion channel modulators for the treatment of pain, epilepsy, and cardiovascular diseases; 2) to study molecular determinants of ion channel modulation and gating. While we have a solid basis for understanding of the physiological role of ion channel function, the molecular mechanisms of ion channel modulation and gating remain elusive. To gain novel insights into atomic scale structural details of ion channel gating, my group is using computational biology approaches to generate experimentally testable structural hypotheses. My group is leading an effort to develop structural models of ion channels with atomic accuracy using computational modeling approaches and use this knowledge for the rational design of novel ion channel modulators that can be used for the treatment of chronic pain, epilepsy, and cardiac arrhythmias. Related recent papers:

https://pubmed.ncbi.nlm.nih.gov/39463944/

https://pubmed.ncbi.nlm.nih.gov/39189871/

https://pubmed.ncbi.nlm.nih.gov/38445990/