Interpretation of results

[cblair8181]

Hi there,

I just ran PyPythia on a matrix of 82 sequences and 7,545,366 bp. The predicted difficulty is 0, yet the proportion of unique topologies is 100%. I would assume that different MP trees would increase the difficulty score, though I understand that this is only one of the features used in the final calculations. Just wanted to make sure that nothing seems out of the ordinary. The full results are below.

Best wishes,
Chris

FEATURES:
num_taxa: 82
num_sites: 7545366
num_patterns: 1750037
num_patterns/num_taxa: 21341.91
num_sites/num_taxa: 92016.66
num_patterns/num_sites: 0.23
proportion_gaps: 0.34
proportion_invariant: 0.95
entropy: 0.03
bollback: -102313008.28
pattern_entropy: 17178759.03
avg_rfdist_parsimony: 0.13
proportion_unique_topos_parsimony: 1.0

[tschuelia]

Hi Chris! Seeing a difficulty of 0.0 for your dataset makes total sense to me. Indeed, every parsimony tree has a different topology, but the topologies are very similar (average pairwise RF-Distance 0.13). Also your dataset has a very high sites-per-taxa and patterns-per-taxa ratio which indicates high signal in your data. So overall it looks like you have a fantastic MSA for performing analyses 🥳

And fun fact regarding the unique topologies: we observed that in the Pythia training data almost all MSAs have a proportion of unique topologies of 1 (see https://doi.org/10.1101/2025.03.25.645182 Figure 2).

I hope this helps 🙂