Accurate Machine Learning Prediction of Protein Circular Dichroism Spectra with Embedded Density Descriptors

Abstract

A data-driven approach to simulate circular dichroism (CD) spectra is appealing for fast protein secondary structure determination, yet the challenge of predicting electric and magnetic transition dipole moments poses a substantial barrier for the goal. To address this problem, we designed a new machine learning (ML) protocol in which ordinary pure geometry-based descriptors are replaced with alternative embedded density descriptors and electric and magnetic transition dipole moments are successfully predicted with an accuracy comparable to first-principle calculation. The ML model is able to not only simulate protein CD spectra nearly 4 orders of magnitude faster than conventional first-principle simulation but also obtain CD spectra in good agreement with experiments. Finally, we predicted a series of CD spectra of the Trp-cage protein associated with continuous changes of protein configuration along its folding path, showing the potential of our ML model for supporting real-time CD spectroscopy study of protein dynamics.

Terahertz Circular Dichroism Spectroscopy of Molecular Assemblies and Nanostructures

“…A wide range of ML methodologies have been proposed to improve computational accuracy, cost, and speed in IR, Raman, VCD, and ECD spectra (Figure ). − Similar advances must be possible for TCD spectroscopy, too. Although there are no (to the best of our knowledge) ML-based methodologies developed and published for TCD yet, special features of this spectroscopy and its implementation for multidimensional analysis of the biological and abiological structures discussed above make this spectroscopy particularly suitable for the application of ML.…”

“…(c) ML-based simulation of ECD spectra of peptides predicted based on 1000 MD configurations. Reprinted with permission from ref . Copyright 2021 Americal Chemical Society.…”

Chemical and Biophysical Signatures of the Protein Corona in Nanomedicine

“…Carbon-based nanomaterials (e.g., carbon nanotubes, graphene, carbon quantum dots, and nanodiamonds), metals (e.g., gold and silver nanoparticles), and metal oxide nanoparticles (e.g., SiO 2 , ZnO, CeO 2 , and TiO 2 nanoparticles) exhibit different corona features driven by corresponding distinct nanoparticle–protein interactions. Specifically, the binding of pristine carbon nanotubes (CNTs) with blood proteins, including bovine fibrinogen, gamma globulin (Ig), Tf, and BSA, showed a positive correlation with the number of aromatic residues, suggesting the binding was governed by π–π stacking between protein aromatic residues (Trp, Phe, and Tyr) and carbon nanotubes in addition to hydrophobic interactions. , The proteins formed well-ordered rodlike structures on the CNT surfaces, with the amphiphilic α-helices possessing a high fraction of hydrophobic residues converted into coil or β-sheet structures. For graphene oxide, the oxygen-containing functional groups (e.g., hydroxyls, epoxides, and carboxyls) enabled the additional contributions of electrostatic attraction and hydrogen bonding to stabilize the protein corona and, consequently, denatured the ordered protein structures into unstructured random coils .…”

Exploring Spectrum‐based Molecular Descriptors for Reaction Performance Prediction

“…Recent years have witnessed a fast development of the usage of spectral information in chemical machine learning. The recognition of protein structures by machine learning methods using spectral information [32] as well as spectrum prediction from protein structure [33,34] have been developed and applied to protein molecular dynamics [35] and real‐time interaction simulations [36] . There are increasing attempts to apply data‐driven ML methods to automate the linking of the spectrum of molecules to structures [37–39] .…”