BMGE

BMGE trims multiple sequence alignments by removing ambiguously aligned, highly variable, or saturated positions to improve phylogenetic inference.

Key Features:

• Entropy scoring: Calculates entropy-like scores for each character (column) in an alignment.
• Similarity-matrix weighting: Weights entropy-like scores using BLOSUM or PAM similarity matrices to distinguish expected versus unexpected variability.
• Block identification and removal: Identifies contiguous sets of characters that fall below a specified threshold score and removes those regions from the alignment.
• Sequence-type support: Accepts DNA, codon, and amino acid alignments.
• Trimming and recoding: Applies trimming and recoding approaches to minimize artifacts from compositional heterogeneity.
• Large-dataset handling: Scales to large datasets derived from numerous complete genomes and semi-automated pipelines where manual curation is limited.
• Phylogenetic accuracy: Produces trimmed datasets that simulation analyses show can lead to more accurate phylogenetic trees, particularly for distantly related sequences.

Scientific Applications:

• Phylogenetic inference refinement: Prepares alignments for tree reconstruction by removing misleading or ambiguous positions.
• Distantly related sequence analysis: Improves signal recovery and tree accuracy when analyzing distantly related sequences.
• Large-scale comparative genomics: Refines alignments from numerous complete genomes processed semi-automatically.
• Mitigating compositional heterogeneity: Reduces phylogenetic artifacts caused by compositional heterogeneity through trimming and recoding.

Methodology:

BMGE computes entropy-like scores per alignment character, weights them with BLOSUM or PAM matrices, identifies contiguous columns below a specified threshold and removes those blocks, and provides trimming and recoding for DNA, codon, or amino-acid alignments.