Cluster Analysis of Coronavirus Sequences using Computational Sequence Descriptors: With Applications to SARS, MERS and SARS-CoV-2 (CoVID-19)

Abstract

Introduction: Coronaviruses comprise a group of enveloped, positive-sense single-stranded RNA viruses that infect humans as well as a wide range of animals. The study was performed on a set of 573 sequences belonging to SARS, MERS and SARS-CoV-2 (CoVID-19) viruses. The sequences were represented with alignment-free sequence descriptors and analyzed with different chemometric methods: Euclidean/Mahalanobis distances, principal component analysis and self-organizing maps (Kohonen networks). We report the cluster structures of the data. The sequences are well-clustered regarding the type of virus; however, some of them show the tendency to belong to more than one virus type.

Background: This is a study of 573 genome sequences belonging to SARS, MERS and SARS-- CoV-2 (CoVID-19) coronaviruses.

Objectives: The aim was to compare the virus sequences, which originate from different places around the world.

Methods: The study used alignment free sequence descriptors for the representation of sequences and chemometric methods for analyzing clusters.

Results: Majority of genome sequences are clustered with respect to the virus type, but some of them are outliers.

Conclusion: We indicate 71 sequences, which tend to belong to more than one cluster.

Keywords: Euclidean distance; MERS; Mahalanobis distance; SARS; SARS-CoV-2 (CoVID-19); alignment-free sequenc descriptors.; clustering; mathematical representation of sequences; principal component analysis.