MathPIP: Classification of Proinflammatory Peptides Using Mathematical Descriptors

Abstract

Proinflammatory peptide (PIP) is a relevant part of the inflammatory response, often the first response of our immune system to strange bodies, i.e., inflammatory-inducing infection, such as COVID-19. Thus, it is essential to have reliable ways to classify and analyze new instances of PIPs. Machine learning (ML) models have been widely employed for the classification of biological sequences, being the basis for most studies in extensive databases of biological information. Most ML algorithms have difficulty to directly deal with these sequences. Thereby, relevant features are extracted from these sequences, making feature extraction one of the key steps in the application of ML algorithms to biological data. Different features have been proposed, many of them based on prior knowledge, such as molecular structures. However, many biological sequences publicly available do not come with prior knowledge. To deal with this limitation, we propose to investigate the use of mathematical descriptors to extract features from PIP sequences. To assess how relevant are the features extracted using mathematical descriptors, we run experiments where we apply three ML algorithms. In these experiments, we obtained a predictive accuracy of 0.7034, which is on par with current PIP classifiers.

J. P. U. Cavalcante, A. C. Gonçalves and R. P. Bonidia—The authors wish it to be known that, in their opinion, the first three authors should be regarded as Joint First Authors.