Like most codon modeling approaches, PhyloCSF assumes statistical independence of each codon site from all the others in the alignment (conditional on the model parameters).
The CRF here models the conditional probability of the alignment given the pair of sequences to be aligned.
The solid y-axis (left) illustrates: cDNA query sequence length (pink circle), P. patula subject sequence length (blue square), conditional reciprocal best BLAST alignment length (gold triangle).
Conditional reciprocal best BLAST alignment results between full-length Sanger sequenced Pinus taeda cDNA and representative Pinus patula transcripts for each cDNA.
THESEUS functions in two modes: (i) one mode for superpositioning structures with identical sequences [such as different (NMR) models] and (ii) an 'alignment mode,' which superpositions structures with different sequences (i.e. with missing data) conditional on a known alignment.
Thus it becomes desirable to propose and investigate more efficient multiple-imputation-like approaches such as using posterior ancestor samples which may offer multiple representative alignment results conditional on a reference species sequences for sequence set probability elicitation.
This algorithm, which we call RNAG, iteratively block samples from the conditional probability distributions P(Structure | Alignment) and P(Alignment | Structure), and in so doing refines the models of both Alignment and Structure.
This algorithm iteratively samples from the conditional probability distributions P(Structure | Alignment) and P(Alignment | Structure).
At each position, the posterior probability distribution of the tree topologies, conditional to the multiple sequence alignment in the window, is computed.
In [ 29], the authors, instead of using the individual tree topologies, first clustered these topologies using a Robinson-Fould distance, and then computed the posterior probability distribution of the clusters conditional to the multiple sequence alignment in the window.
Tabei and Asai [ 45] propose an approach SCARNA_LM where they model alignments using a conditional random field (CRF).
