A gene tree built from your alignment gives an unexpected topology. Which alignment problems would you check?

Why

An unexpected tree can be real biology, but alignment problems are more common causes. First, the input. If a paralog sits among orthologs, the tree mixes a duplication with speciation and species appear in the wrong place, the error the OrthoFinder authors describe for best-hit methods.

Fragments and contaminant sequences also draw odd branches. Second, the alignment itself. The Clustal Omega paper notes that progressive alignment is greedy, so an early mistake is never corrected, which can misplace whole blocks in divergent members.

Inspecting known motifs often reveals this. Third, the columns. Poorly aligned, gap-rich regions add noise, and the trimAl authors note that removing poorly aligned regions improves later analyses.

Trimming choices can change topology, so test more than one. A useful robustness check is to rebuild the tree from alignments made with a different aligner and different trimming. Splits that persist across choices deserve belief.

Splits that appear with only one combination are artifacts until shown otherwise, and low support values should be reported as unresolved rather than interpreted.