SATe reading: inferable branches and alignment error

SATe reading — zero-event branches get contracted for PIMT. Interesting metric question: why penalize a branch distinction the sequence data never supported? keep the original tree too.

paper notes: Liu et al., SATe, 2009. The presentation describes an iterative heuristic alternating alignment proposals and maximum-likelihood tree estimation, using center-tree decomposition to propose a new alignment.

evaluation design summarized in the slides: ROSE simulation, thirty evolutionary models and twenty replicates per model, at 500- and 1,000-taxon scales; 1,000-base root sequences; biological data analysis alongside simulation. These are the cited paper's experiments, not new ReconBench runs.

SATe reading notes: PIMT and zero-event branches, from Liu et al. (2009).

SATe reading notes: SP-FN alignment error, from Liu et al. (2009).

metric idea: compare against a potentially inferable model tree (PIMT), formed by contracting branches with zero evolutionary events, rather than penalizing a method for failing to recover unsupported distinctions. Alignment SP-FN counts true homologous pairs absent from the estimate.

review questions: which simulated branches have enough event evidence to be inferable, and how is that condition encoded in the reference? Keep the original true tree and contracted evaluation tree as separate artifacts. This reading is useful for designing fair accuracy metrics, but does not replace the actual benchmark protocol.