Rescuing concatenation with maximum likelihood using supermatrix rooted triples
Concatenated alignments are often used to infer species-level reslationships. Previous studies have shown that analysis of concatenated alignments using maximum likelihood (ML) can produce misleading results. We develop a polynomial-time method that constructs a species tree through inferred rooted triples from concatenated alignments. We call this method SuperMatrix Rooted Triple (SMRT). We show that SMRT performs well in simulations and then show that it is a statistically consistent estimator of a clocklike species tree under a binary substitution model as well as other assumptions. SMRT is therefore a computationally efficient and statistically consistent estimator of species trees.