Enabling Rational Design of Drug Targeting Protein-Protein Interactions with Physics-based Computational Modeling

Project Summary
Enabling Rational Design of Drug Targeting Protein-Protein Interactions with Physics-based
Computational Modeling.
Current drugs mainly target single proteins; future drugs will modulate the interaction between proteins. Protein-
protein interactions (PPIs) are the building blocks of the complex interaction network that regulates cells' and
viruses' behaviors and life cycles. PPI-targeting drugs' ability to modulate these networks will provide new tools
to fight cancer and genetic diseases, and provide new classes of antibiotics and antivirals. PPI-targeting drugs
will act on the imbalance of cells' PPI network caused by cancer-related genetic mutations and block bacterial
and viral infections by disrupting PPIs essential for the progression of the infection. To rationally design this new
class of drugs, we need to develop computational tools to predict the strength of the interactions between two
proteins, and the effects of mutations and drugs on those interactions. Machine learning and artificial intelligence
techniques are the basis on which many drug design tools are developed. These techniques rely on databases
that can be used to train the models. However, such databases don't exist for PPIs, and might be impossible to
build due to the uniqueness of PPI interfaces. Physics-based methods, like molecular dynamics simulations,
offer a principled way to develop drug-design tools without relying on the existence of databases. Although the
conformational space of biologically relevant systems is typically too vast to be sampled effectively using physics-
based simulation techniques, the Modelling Employing Limited Data (MELD) method can overcome this limitation
by using external information. MELD has been successfully used to fold proteins, predict drug binding affinities,
and predict the structure of protein dimers. In this grant, we propose to leverage the MELD method to create
some of the computational tools that are currently missing to design PPI-targeting drugs. In Aim 1, we propose
to develop a protocol to quantify the effect that mutations have on proteins' ability to interact. In Aim 2, we
propose to develop computational tools to screen drugs based on their effect on PPIs. The benchmark for
developing our tools will be small biological systems that have been previously studied using other physics-
based approaches. The testing bed for our tools will be the calculation of key properties of biologically relevant
size systems that have been studied experimentally, that are available in databases, and that are too big for any
currently available computational tools to tackle. The final test for our tools will be the prediction of key properties
on systems that our experimental collaborators will help us investigate. At the end of this project, we will have
developed and tested the tools to fill the current gap in the rational design of PPI-targeting drugs. In the long
term, these tools will allow us to understand the molecular mechanisms of cancer and genetic diseases, and will
help the rational design of the next generation of drugs to treat cancer and viral and bacterial infections. These
are drugs that will target PPI interactions rather than single proteins. The tools and the knowledge we will acquire
from this grant will be the stepping stone for the future research of our group, which aims to study the molecular
mechanisms of genetic diseases, including cancer, and design antiviral and antibiotic drugs.

Grant Number: 5R16GM150512-03
NIH Institute/Center: NIH
Principal Investigator: Emiliano Brini