Improved regression model to predict an impact of SOD1 mutations on ALS patients survival time based on analysis of hydrogen bond stability.

Amyotrophic lateral sclerosis (ALS) is an incurable neurodegenerative disease characterised by the inevitable degeneration of central and peripheral motor neurons. Aggregation of mutant SOD1 is one of the molecular mechanisms underlying the onset of the disease. There are a number of regression models designed to predict the survival of patients based on an analysis of experimental data on thermostability, heterodimerisation energy, and changes in the hydrophobicity of SOD1 mutants. Previously, we proposed regression models linking the change in the stability of hydrogen bonds in mutant SOD1 calculated using molecular dynamics and elastic networks with patients survival time. In this study, these models were improved in terms of accuracy of survival time prediction by taking into account the variance of survival time values relative to the mean, the number of patients carrying each specific mutation, and the use of random forest regression as a regression method. The accuracy of the previous models was roughly 5.2 years while the accuracy of the new ones are up to 4 years. The model is also superior to those published by other authors. It was found that the hydrogen bonds important for prediction of survival time are formed by residues at positions located in the regions of the protein responsible for aggregation as well as in structural and functionally important sites.