MuscleAlign follows the three stages of MUSCLE (Edgar 2004): a draft
from k-mer distances (1 - F, F the shared k-mer count over
min(|x|, |y|) - k + 1), a UPGMA guide tree and progressive
profile-profile alignment (Gotoh affine gaps, frequency-weighted mean
substitution score per column pair); an improved alignment from Kimura
distances -ln(1 - p - p^2/5) of the draft (10 when undefined) and a
second UPGMA tree; and tree-dependent restricted partitioning, re-aligning
the two sub-alignments on either side of each edge (deepest first) and
keeping the result when the sum-of-pairs score rises, until a pass brings
no gain or max_iters. Simplifications against MUSCLE 3: no sequence
weights, constant gap penalties (ends included) and BLOSUM62 or +5/-4
nucleotide scores instead of the log-expectation profile function.
SumOfPairsScore is the objective: over all row pairs, columns gapped
in both are dropped, residue pairs score S(a, b) and a gap run of
length L costs gap_open + (L - 1) gap_extend. Identical to
the Python arm morie.fn.musclemsa.
Usage
MuscleAlign(
sequences,
matrix = NULL,
gap_open = 10,
gap_extend = 1,
k = 3,
max_iters = 16
)
SumOfPairsScore(alignment, matrix = NULL, gap_open = 10, gap_extend = 1)Arguments
- sequences
Character vector of unaligned sequences.
- matrix
"blosum62","nucleotide"or a square score matrix with letter dimnames; chosen from the alphabet whenNULL.- gap_open, gap_extend
Affine gap costs.
- k
k-mer length for the draft distances.
- max_iters
Maximum refinement passes.
- alignment
Character vector of aligned rows (gap
-).
Value
MuscleAlign: list with alignment (input order),
sp_score, sp_draft, sp_improved, accepted,
tree1 and tree2 (UPGMA merges, 1-based node ids).
SumOfPairsScore: the score.