Using temperature coefficients to support resonance assignment of intrinsically disordered proteins.

The resonance assignment of large intrinsically disordered proteins (IDPs) is difficult due to the low dispersion of chemical shifts (CSs). Luckily, CSs are often specific for certain residue types, which makes the task easier. Our recent work showed that the CS-based spin-system classification can be improved by applying a linear discriminant analysis (LDA). In this paper, we extend a set of classification parameters by adding temperature coefficients (TCs), i.e., rates of change of chemical shifts with temperature. As demonstrated previously by other groups, the TCs in IDPs depend on a residue type, although the relation is often too complex to be predicted theoretically. Thus, we propose an approach based on experimental data; CSs and TCs values of residues assigned using conventional methods serve as a training set for LDA, which then classifies the remaining resonances. The method is demonstrated on a large fragment (1-239) of highly disordered protein Tau. We noticed that adding TCs to sets of chemical shifts significantly improves the recognition efficiency. For example, it allows distinguishing between lysine and glutamic acid, as well as valine and isoleucine residues based on ${\text{H}}^{\text{N}}$ , N, ${\text{C}}_{\alpha }$ and C $_{}^{\text{'}}$ data. Moreover, adding TCs to CSs of ${\text{H}}^{\text{N}}$ , N, ${\text{C}}_{\alpha }$ , and C $_{}^{\text{'}}$ is more beneficial than adding ${\text{C}}_{\beta }$ CSs. Our program for LDA analysis is available at https://github.com/gugumatz/LDA-Temp-Coeff .