Alignments and Trimmed Alignments: Their Characteristics and Phylogenetic Trees

The multiple sequence alignment is important in bioinformatics. In this study, we used (1) MAFFT, MUSCLE, Clustal W, and Clustal X with substitution matrices, BLOSUM, Gonnet and PAM, to align 3228 coronavirus spike proteins; (2) the trimAl to trim the alignments using automated1, gappyout, strict and strictplus; (3) ME, ML, MP, NJ, and UPGMA to untrimmed and trimmed alignment to construct 140 phylogenetic trees. We tested the results using (a) amino acid composition, (b) alignment statistics, (c) alignment order and positions regarding referenced sequence, (d) phylogenetic information of alignment, (e) Robinson-Foulds distance, (f) locality of referenced sequence across phylogenetidc trees, and (g) effects of different methods with the model II one-way ANOVA. The results show (i) the alignment algorithms mainly affect the order of sequences in alignment, (ii) MUSCLE has the longest alignment, (iii) substitution matrices likely affect the alignment length under the same alignment algorithm, (iv) automated1 and strict generate the same result, (v) the longer the alignment, the heavier the trimming, (vi) trimming keeps the phylogenetic information, (vii) BLOSUM and Gonnet likely construct phylogenetic trees with zero Robinson-Foulds distance, (viii) the locality of referenced sequence in branch level varies across phylogenetic trees, and (ix) trimming generate identical sequences.