def guess_amplicons(fastq_r1,fastq_r2,number_of_reads_to_consider,flash_command,max_paired_end_reads_overlap,min_paired_end_reads_overlap,aln_matrix,needleman_wunsch_gap_open,needleman_wunsch_gap_extend,min_freq_to_consider=0.2,amplicon_similarity_cutoff=0.95):
"""
guesses the amplicons used in an experiment by examining the most frequent read (giant caveat -- most frequent read should be unmodified)
input:
fastq_r1: path to fastq r1 (can be gzipped)
fastq_r2: path to fastq r2 (can be gzipped)
number_of_reads_to_consider: number of reads from the top of the file to examine
flash_command: command to call flash
min_paired_end_reads_overlap: min overlap in bp for flashing (merging) r1 and r2
max_paired_end_reads_overlap: max overlap in bp for flashing (merging) r1 and r2
needleman_wunsch_gap_open: alignment penalty assignment used to determine similarity of two sequences
needleman_wunsch_gap_extend: alignment penalty assignment used to determine similarity of two sequences
min_freq_to_consider: selected ampilcon must be frequent at least at this percentage in the population
amplicon_similarity_cutoff: if the current amplicon has similarity of greater than this cutoff to any other existing amplicons, it won't be added
returns:
list of putative amplicons
"""
seq_lines = get_most_frequent_reads(fastq_r1,fastq_r2,number_of_reads_to_consider,flash_command,max_paired_end_reads_overlap,min_paired_end_reads_overlap)
curr_amplicon_id = 1
amplicon_seq_arr = []
#add most frequent amplicon to the list
count,seq = seq_lines[0].strip().split()
amplicon_seq_arr.append(seq)
curr_amplicon_id += 1
#for the remainder of the amplicons, test them before adding
for i in range(1,len(seq_lines)):
count,seq = seq_lines[i].strip().split()
last_count,last_seq = seq_lines[i-1].strip().split()
#if this allele is present in at least XX% of the samples
if float(last_count)/float(number_of_reads_to_consider) > min_freq_to_consider:
this_amplicon_seq_arr = amplicon_seq_arr[:]
this_amplicon_max_pct = 0 #keep track of similarity to most-similar already-found amplicons
for amp_seq in this_amplicon_seq_arr:
ref_incentive = np.zeros(len(amp_seq)+1,dtype=np.int)
fws1,fws2,fwscore=CRISPResso2Align.global_align(seq,amp_seq,matrix=aln_matrix,gap_incentive=ref_incentive,gap_open=needleman_wunsch_gap_open,gap_extend=needleman_wunsch_gap_extend,)
rvs1,rvs2,rvscore=CRISPResso2Align.global_align(reverse_complement(seq),amp_seq,matrix=aln_matrix,gap_incentive=ref_incentive,gap_open=needleman_wunsch_gap_open,gap_extend=needleman_wunsch_gap_extend,)
#if the sequence is similar to a previously-seen read, don't add it
min_len = min(len(last_seq),len(seq))
max_score = max(fwscore,rvscore)
if max_score/float(min_len) > this_amplicon_max_pct:
this_amplicon_max_pct = max_score/float(min_len)
#if this amplicon was maximally-similar to all other chosen amplicons by less than amplicon_similarity_cutoff, add to the list
if this_amplicon_max_pct < amplicon_similarity_cutoff:
amplicon_seq_arr.append(seq)
curr_amplicon_id += 1
else:
break
return amplicon_seq_arr
def RunCRISPResso2(self, strQuerySeqAfterBarcode, strRefSeqAfterBarcode,
npGapIncentive):
listResult = CRISPResso2Align.global_align(
strQuerySeqAfterBarcode.upper(),
strRefSeqAfterBarcode.upper(),
matrix=self.npAlnMatrix,
gap_open=self.floOg,
gap_extend=self.floOe,
gap_incentive=npGapIncentive)
return listResult
def test_global_align():
"""General alignment tests."""
seq1, seq2, score = CRISPResso2Align.global_align('ATTA',
'ATTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 0],
dtype=int))
assert seq1 == 'ATTA'
assert seq2 == 'ATTA'
assert score == 100
def guess_guides(amplicon_sequence,
fastq_r1,
fastq_r2,
number_of_reads_to_consider,
flash_command,
max_paired_end_reads_overlap,
min_paired_end_reads_overlap,
aln_matrix,
needleman_wunsch_gap_open,
needleman_wunsch_gap_extend,
min_edit_freq_to_consider=0.1,
pam_seq="NGG",
min_pct_subs_in_base_editor_win=0.8):
"""
guesses the guides used in an experiment by identifying the most-frequently edited positions, editing types, and PAM sites
input:
ampilcon_sequence - amplicon to analyze
fastq_r1: path to fastq r1 (can be gzipped)
fastq_r2: path to fastq r2 (can be gzipped)
number_of_reads_to_consider: number of reads from the top of the file to examine
flash_command: command to call flash
min_paired_end_reads_overlap: min overlap in bp for flashing (merging) r1 and r2
max_paired_end_reads_overlap: max overlap in bp for flashing (merging) r1 and r2
needleman_wunsch_gap_open: alignment penalty assignment used to determine similarity of two sequences
needleman_wunsch_gap_extend: alignment penalty assignment used to determine similarity of two sequences
min_edit_freq_to_consider: edits must be at least this frequency for consideration
pam_seq: pam sequence to look for (can be regex or contain degenerate bases)
min_pct_subs_in_base_editor_win: if at least this percent of substitutions happen in the predicted base editor window, return base editor flag
returns:
tuple of (putative guide, boolean is_base_editor)
or (None, None)
"""
seq_lines = get_most_frequent_reads(fastq_r1, fastq_r2,
number_of_reads_to_consider,
flash_command,
max_paired_end_reads_overlap,
min_paired_end_reads_overlap)
amp_len = len(amplicon_sequence)
gap_incentive = np.zeros(amp_len + 1, dtype=np.int)
include_idxs = set(range(0, amp_len))
all_indel_count_vector = np.zeros(amp_len)
all_sub_count_vector = np.zeros(amp_len)
tot_count = 0
for i in range(len(seq_lines)):
count, seq = seq_lines[i].strip().split()
count = int(count)
tot_count += count
fws1, fws2, fwscore = CRISPResso2Align.global_align(
seq,
amplicon_sequence,
matrix=aln_matrix,
gap_incentive=gap_incentive,
gap_open=needleman_wunsch_gap_open,
gap_extend=needleman_wunsch_gap_extend,
)
payload = CRISPRessoCOREResources.find_indels_substitutions(
fws1, fws2, include_idxs)
all_indel_count_vector[payload['all_insertion_positions']] += count
all_indel_count_vector[payload['all_deletion_positions']] += count
all_sub_count_vector[payload['all_substitution_positions']] += count
max_loc = np.argmax(all_indel_count_vector)
max_val = all_indel_count_vector[max_loc]
#return nothing if the max edit doesn't break threshold
if max_val / float(tot_count) < min_edit_freq_to_consider:
return (None, None)
pam_regex_string = pam_seq.upper()
pam_regex_string = pam_regex_string.replace('I', '[ATCG]')
pam_regex_string = pam_regex_string.replace('N', '[ATCG]')
pam_regex_string = pam_regex_string.replace('R', '[AG]')
pam_regex_string = pam_regex_string.replace('Y', '[CT]')
pam_regex_string = pam_regex_string.replace('S', '[GC]')
pam_regex_string = pam_regex_string.replace('W', '[AT]')
pam_regex_string = pam_regex_string.replace('K', '[GT]')
pam_regex_string = pam_regex_string.replace('M', '[AC]')
pam_regex_string = pam_regex_string.replace('B', '[CGT]')
pam_regex_string = pam_regex_string.replace('D', '[AGT]')
pam_regex_string = pam_regex_string.replace('H', '[ACT]')
pam_regex_string = pam_regex_string.replace('V', '[ACG]')
is_base_editor = False
#offset from expected position
for offset in (0, +1, -1, +2, +3, +4, -2):
#forward direction
#find pam near max edit loc
pam_start = max_loc + 4 + offset
pam_end = max_loc + 7 + offset
guide_start = max_loc - 16 + offset
guide_end = max_loc + 4 + offset
base_edit_start = max_loc - 16 + offset
base_edit_end = max_loc - 6 + offset
if pam_start > 0 and guide_end < amp_len:
if re.match(pam_regex_string,
amplicon_sequence[pam_start:pam_end]):
guide_seq = amplicon_sequence[guide_start:guide_end]
sum_base_edits = sum(
all_sub_count_vector[base_edit_start:base_edit_end])
#if a lot of edits are in the predicted base editor window, set base editor true
#specifically, if at least min_pct_subs_in_base_editor_win % of substitutions happen in the predicted base editor window
if sum_base_edits > min_pct_subs_in_base_editor_win * sum(
all_sub_count_vector):
is_base_editor = True
return (guide_seq, is_base_editor)
#reverse direction
pam_start = max_loc - 5 - offset
pam_end = max_loc - 2 - offset
guide_start = max_loc - 2 - offset
guide_end = max_loc + 18 - offset
base_edit_start = max_loc + 8 - offset
base_edit_end = max_loc + 18 - offset
if pam_start > 0 and guide_end < amp_len:
if re.match(pam_regex_string,
amplicon_sequence[pam_start:pam_end]):
guide_seq = amplicon_sequence[guide_start:guide_end]
sum_base_edits = sum(
all_sub_count_vector[base_edit_start:base_edit_end])
#if a lot of edits are in the predicted base editor window, set base editor true
#specifically, if at least min_pct_subs_in_base_editor_win % of substitutions happen in the predicted base editor window
if sum_base_edits > min_pct_subs_in_base_editor_win * sum(
all_sub_count_vector):
is_base_editor = True
return (guide_seq, is_base_editor)
return (None, None)
def test_global_align_gap_incentive_s1():
"""Test the global_align gap incentives for gaps in sequence 1 (the first sequence)."""
seq1, seq2, score = CRISPResso2Align.global_align('ATTTA',
'ATTTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 0, 0],
dtype=int))
# print('seq1: ' + seq1 + ' seq2: ' + seq2 + ' score ' + str(score))
assert seq1 == 'ATTTA'
assert seq2 == 'ATTTA'
assert score == 100
seq1, seq2, score = CRISPResso2Align.global_align('ATTTA',
'ATTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[1, 0, 0, 0, 0],
dtype=int))
assert seq1 == 'ATTTA'
assert seq2 == 'ATT-A'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTTA',
'ATTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 1, 0, 0, 0],
dtype=int))
assert seq1 == 'ATTTA'
assert seq2 == 'A-TTA'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTTA',
'ATTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 1, 0, 0],
dtype=int))
assert seq1 == 'ATTTA'
assert seq2 == 'AT-TA'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTTA',
'ATTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 1, 0],
dtype=int))
assert seq1 == 'ATTTA'
assert seq2 == 'ATT-A'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTTA',
'ATTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 1, 0],
dtype=int))
assert seq1 == 'ATTTA'
assert seq2 == 'ATT-A'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTTA',
'ATTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 1],
dtype=int))
assert seq1 == 'ATTTA'
assert seq2 == 'ATT-A'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTTT',
'TTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 0],
dtype=int))
assert seq1 == 'TTTTT'
assert seq2 == 'TTTT-'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTTT',
'TTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[1, 0, 0, 0, 0],
dtype=int))
assert seq1 == 'TTTTT'
assert seq2 == '-TTTT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTTT',
'TTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 1, 0, 0, 0],
dtype=int))
assert seq1 == 'TTTTT'
assert seq2 == 'T-TTT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTTT',
'TTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 1, 0, 0],
dtype=int))
assert seq1 == 'TTTTT'
assert seq2 == 'TT-TT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTTT',
'TTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 1, 0],
dtype=int))
assert seq1 == 'TTTTT'
assert seq2 == 'TTT-T'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTTT',
'TTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 1],
dtype=int))
assert seq1 == 'TTTTT'
assert seq2 == 'TTTT-'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
def test_global_align_gap_incentive_s2():
"""Test the global_align gap incentives for gaps in sequence 2 (the second sequence)."""
seq1, seq2, score = CRISPResso2Align.global_align('ATTA',
'ATTTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[1, 0, 0, 0, 0, 0],
dtype=int))
assert seq1 == 'ATT-A'
assert seq2 == 'ATTTA'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTA',
'ATTTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 1, 0, 0, 0, 0],
dtype=int))
assert seq1 == 'A-TTA'
assert seq2 == 'ATTTA'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTA',
'ATTTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 1, 0, 0, 0],
dtype=int))
assert seq1 == 'AT-TA'
assert seq2 == 'ATTTA'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTA',
'ATTTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 1, 0, 0],
dtype=int))
assert seq1 == 'ATT-A'
assert seq2 == 'ATTTA'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTA',
'ATTTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 1, 0],
dtype=int))
assert seq1 == 'ATT-A'
assert seq2 == 'ATTTA'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('ATTA',
'ATTTA',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 0, 1],
dtype=int))
assert seq1 == 'ATT-A'
assert seq2 == 'ATTTA'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTT',
'TTTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[1, 0, 0, 0, 0, 0],
dtype=int))
assert seq1 == '-TTTT'
assert seq2 == 'TTTTT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTT',
'TTTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 1, 0, 0, 0, 0],
dtype=int))
assert seq1 == 'T-TTT'
assert seq2 == 'TTTTT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTT',
'TTTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 1, 0, 0, 0],
dtype=int))
assert seq1 == 'TT-TT'
assert seq2 == 'TTTTT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTT',
'TTTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 1, 0, 0],
dtype=int))
assert seq1 == 'TTT-T'
assert seq2 == 'TTTTT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTT',
'TTTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 1, 0],
dtype=int))
assert seq1 == 'TTTT-'
assert seq2 == 'TTTTT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)
seq1, seq2, score = CRISPResso2Align.global_align('TTTT',
'TTTTT',
matrix=ALN_MATRIX,
gap_incentive=np.array(
[0, 0, 0, 0, 0, 1],
dtype=int))
assert seq1 == 'TTTT-'
assert seq2 == 'TTTTT'
assert round(score, 3) == round(100 * 4 / 5.0, 3)