RVD similarity matrix for aligning RVD sequences — .rvd_score

The domain-level similarity matrix is keyed by dom_code, which is meaningless for an RVD sequence, so RVD alignments have had no scoring matrix at all – MAFFT treated NI and NN as no more alike than NI and HD. This supplies the missing one, from the Spearman correlation of each RVD's A/C/G/T preference profile (rvdSimDf, derived from TALVEZ's mat1).

Usage

.rvd_score_table(residues)

Arguments

residues: Character vector of the RVDs present in the alignment.

Value

A data frame of id1, id2, sim covering every ordered pair of residues.

Details

XX – a terminus detected but not identified – has a uniform base profile, so its correlation with everything is undefined. Those cells are filled as neutral (0): we know nothing about it, so it should neither attract nor repel.

The one exception is XX against itself, which is set to the maximum. That is forced, not chosen: MAFFT produces unusable output when the diagonal is not high, and a low diagonal would anyway assert that an XX must not align with an XX, which is a stronger claim than ignorance.

Scale is irrelevant – MAFFT normalises the matrix, so a linear rescale or offset leaves the alignment unchanged. Only relative structure matters, so the correlations are used as they are.