Modeling Site-and-Branch-Heterogeneity with GFmix
Bibliographic record
Abstract
Phylogenetic trees are often inferred from protein sequences sampled from diverse taxa across the tree of life. The compositions of these amino acid sequences may be heterogeneous across both sites and branches, particularly if deep phylogenetic divergences are the focus. Under some conditions, failure to model this compositional heterogeneity can lead to phylogenetic artefacts. However, the computational cost of phylogenetic inference with models accounting for compositional heterogeneity can be prohibitive. The originally proposed site-and-branch-heterogeneous GFmix model accounts for changing relative frequencies of G, A, R, and P (GARP) vs. F, Y, M, I, N, and K (FYMINK) amino acids resulting from extreme variation in G+C content among taxa. This GFmix model modifies a fitted site-heterogeneous profile mixture model in a branch-specific manner using parameters that reflect branch-specific amino acid compositions. This approach has been shown to improve likelihoods and reduce compositional artefacts. However, the original implementation of the model includes constraints which may sacrifice accuracy for computability and is limited to modeling variation in GARP/FYMINK composition. Here we investigate the properties of the original GFmix model in greater depth and present several improvements to the model. The improved GFmix models permit fewer constraints on branch-specific composition parameters, allow modeling of user-defined compositional heterogeneity, and provide for full maximum-likelihood optimization of parameters. We have also developed new methods for detecting compositional heterogeneity directly from sequence data. Analyses of simulated site-and-branch-heterogeneous data indicates that the improved GFmix models better estimate branch-specific compositions and branch lengths in heterogeneous trees. We applied the various versions of the GFmix model to a real dataset with known compositional heterogeneity artefacts. We find that the most complex GFmix model with full maximum likelihood parameter optimization consistently supports the correct tree over the artefactual tree with improved likelihoods. All implementations of the GFmix model and related scripts are available from https://www.mathstat.dal.ca/~tsusko/software.html.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".