Integration of Convolutional Neural Networks in the Analysis of GRB2 Protein Structures using SAXS

• Grant number: 25/07351-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: July 01, 2025
• End date: June 30, 2026
• Field of knowledge: Biological Sciences - Biophysics - Molecular Biophysics
• Principal Investigator: Fernando Alves de Melo
• Grantee: Guilherme Dias Fusari
• Host Institution: Instituto de Biociências, Letras e Ciências Exatas (IBILCE). Universidade Estadual Paulista (UNESP). Campus de São José do Rio Preto. São José do Rio Preto , SP, Brazil
• Associated research grant: 23/01744-5 - Correlation between the structure and dynamics of GRB2 protein and its folding excited states, AP.R

Abstract

The Small-Angle X-ray Scattering (SAXS) technique plays a crucial role in structural biology, particularly in the study of macromolecules such as proteins and protein complexes in solution. Its primary importance lies in its ability to provide information about the conformation, size, and dynamics of molecules under conditions that mimic the natural physiological environment. Unlike other methodologies, SAXS allows the analysis of samples in their native state, without the need for crystallization or environmental restrictions, meaning it can capture the actual conformation of proteins in solution. In addition to its capacity to provide scattering profiles that reveal details about low-resolution structures, SAXS is also a rapid and relatively straightforward technique to perform, making it accessible for many laboratories. Recent advancements in computing, particularly in the integration of artificial intelligence tools, have generated significant new opportunities for the SAXS technique. Innovative approaches, such as AlphaFold, which employs neural networks to predict protein structures from amino acid sequences, demonstrate enormous potential when combined with SAXS data. This synergy can not only enhance the accuracy of structural predictions but also provide a more comprehensive understanding of the molecular dynamics involved in biological systems, allowing for deeper insights into how proteins function in their native environment. Focusing on this, the objective of our work is to develop a methodology to evaluate the quality of the fit of protein structures in dummy atom models, integrating the use of convolutional neural networks with SAXS data obtained in our laboratory for the GRB2 protein. This protein is of extreme biological importance as it participates in crucial cellular signaling pathways and, structurally, exhibits flexibility that allows for the exploration of a broad range of conformational configurations. This process will function as a two-way street: experimental data will feed the model, enabling its continuous refinement, while the resulting models will contribute to complement and enrich the ongoing experimental results. (AU)