Prediction of the amount of secondary structure of proteins using unassigned NMR spectra: a tool for target selection in structural proteomics

Abstract

With the advent of structural genomics, the need for fast structural information about unknown proteins has increased. We describe a new methodology, based on 13 C, 15 N and 1 H chemical shift dispersion to predict the amount of secondary structure of unassigned proteins from their 15 N-and/or 13 C-edited heteronuclear single quantum coherence (HSQC) spectra. This methodology has been coded into a software called PASSNMR (Prediction of the Amount of Secondary Structure by Nuclear Magnetic Resonance), which can be accessed directly from the Internet. PASSNMR program is a powerful tool for screening proteins for proteomic or structural genomic investigations when used with recent methodologies that take advantage of the use of the antibiotic rifampicin to selectively label the heterologous proteins expressed in E. coli. PASSNMR analysis can be useful as a first approach to predict the amount of secondary structure in proteins to structural genomics. Information about the secondary structure of proteins can be obtained even before protein purification, with small quantities of protein, just by performing two simple nuclear magnetic resonance (NMR) experiments and using PASSNMR program.

Characterization of protein secondary structure from NMR chemical shifts

“…Recently, Moreau et al, [53] have presented a method similar to the above mentioned approach using heteronuclear ( 13 C and 15 N) chemical shifts in addition to the 1 H chemical shifts. This method is called PASSNMR (Prediction of the Amount of Secondary Structure by Nuclear Magnetic Resonance).…”

“…The goal of this approach and many of the other methods discussed in this article is to predict the amount of secondary structure in proteins for structural genomics applications. Overall reliability of the PASSNMR approach for helical, sheet and coiled structures are 72%, 74% and 42%, respectively and the using only the 15 N- 1 H data, these values drop to 49% for helix and 50% for sheets [53]. …”

HSQC2STRUC: A Machine Learning Model for Protein Secondary Structure Prediction using Unassigned NMR Spectra

“…The two most recent approaches in predicting secondary structure content with NMR data are TALOS+/ TALOS-N [26; 25] and PASSNMR [22]. TALOS is widely used in the scientific community, while PASSNMR is no longer publicly available but formed the basis for the here-developed model.…”

“…This renders the expression of protein expensive and the overall process time-consuming. First attempts to use unassigned 1 H, 15 N-HSQC spectra to estimate secondary structure content served as a prove-of-concept [22]. Still, it was not competitive with other tools due to the limitation of a small data set of 64 proteins and the utilization of a simple linear regression model.…”