biostudies-literatureWaeytens JFonds De La Recherche Scientifique - FNRSService Public de Wallonie621-627https://www.ebi.ac.uk/biostudies/studies/S-EPMC9851152biostudies-literatureUnknown95(2)Nanoscale infrared spectroscopy (AFMIR) is becoming an important tool for the analysis of biological sample, in particular protein assemblies, at the nanoscale level. While the amide I band is usually used to determine the secondary structure of proteins in Fourier transform infrared spectroscopy, no tool has been developed so far for AFMIR. The paper introduces a method for the study of secondary structure of protein based on a protein library of 38 well-characterized proteins. Ascending stepwise linear regression (ASLR) and partial least square (PLS) regression were used to correlate spectrum characteristic bands with the major secondary structures (α-helixes and β-sheets). ASLR appears to provide better results than PLS. The secondary structure predictions are characterized by a root mean square standard error in a cross validation of 6.39% for α-helixes and 6.23% for β-sheets.Analytical chemistryDetermination of Secondary Structure of Proteins by Nanoinfrared Spectroscopy.PMC9851152DGO630467715O001518FRaussens VDazzi AGoormaghtigh EDe Meutter JWaeytens JfalseDetermination of Secondary Structure of Proteins by Nanoinfrared Spectroscopy.Nanoscale infrared spectroscopy (AFMIR) is becoming an important tool for the analysis of biological sample, in particular protein assemblies, at the nanoscale level. While the amide I band is usually used to determine the secondary structure of proteins in Fourier transform infrared spectroscopy, no tool has been developed so far for AFMIR. The paper introduces a method for the study of secondary structure of protein based on a protein library of 38 well-characterized proteins. Ascending stepwise linear regression (ASLR) and partial least square (PLS) regression were used to correlate spectrum characteristic bands with the major secondary structures (α-helixes and β-sheets). ASLR appears to provide better results than PLS. The secondary structure predictions are characterized by a root mean square standard error in a cross validation of 6.39% for α-helixes and 6.23% for β-sheets.2023-01-01T00:00:00Z2023 Jan2025-04-26T07:02:23.02Z2025-04-06T12:03:38.06ZS-EPMC98511523659892910.1021/acs.analchem.2c01431