def test_alignment_std_local(err):
A = Alphabet('ACGT')
M = MutationProcess(A, subst_probs=err, go_prob=err, ge_prob=err)
subst_scores, (go_score, ge_score) = M.log_odds_scores()
S = rand_seq(A, 100)
T, tx = M.mutate(S)
T = A.parse('A' * 100) + T + A.parse('G' * 100)
mutation_aln = Alignment(S, T, tx)
mutation_score = mutation_aln.calculate_score(subst_scores, go_score,
ge_score)
aligner = Aligner(S, T, subst_scores=subst_scores, go_score=go_score,
ge_score=ge_score, alnmode=STD_MODE, alntype=LOCAL)
with aligner:
reported_score = aligner.solve()
assert round(reported_score, 3) >= round(mutation_score, 3), \
'optimal alignment scores better than the known transcript'
alignment = aligner.traceback()
aln_score = alignment.calculate_score(subst_scores, go_score, ge_score)
assert round(aln_score, 3) == round(reported_score, 3), \
'The alignment score should be calculated correctly'
ori_len = Alignment.projected_len(alignment.transcript, on='origin')
mut_len = Alignment.projected_len(alignment.transcript, on='mutant')
assert ori_len <= len(S) and mut_len < len(T), \
'Local alignments do not cover the entirety of both sequences'
def test_log_odds_scores():
A = Alphabet('ACGT')
# linear gap model
P = MutationProcess(A, subst_probs=.1, ge_prob=.1, go_prob=.1)
subst_scores, (go_score, ge_score) = P.log_odds_scores()
assert go_score == 0. and ge_score < 0
match_pos = [(i, i) for i in range(len(A))]
mismatch_pos = [(i, j) for i, j in combinations(range(len(A)), 2)]
assert all(subst_scores[i][j] < 0 for i, j in mismatch_pos)
assert all(subst_scores[i][j] > 0 for i, j in match_pos)
# affine gap model
P = MutationProcess(A, subst_probs=.1, ge_prob=.2, go_prob=.1)
subst_scores, (go_score, ge_score) = P.log_odds_scores()
assert ge_score < 0
# do mismatch scores go down if subst probs are decreased?
P = MutationProcess(A, subst_probs=.01, ge_prob=.2, go_prob=.1)
new_subst_scores, _ = P.log_odds_scores()
assert new_subst_scores[0][1] < subst_scores[0][1], \
'mismatch scores become more negative with lower mismatch probs'
