Skip to content

Devel #10

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
sfilges merged 2 commits into from
Mar 17, 2026
Merged

Devel #10

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0ab70c0
Added _BLAST_DTYPES and _BLAST_CATEGORICAL_COLS constants to improve …
sfilges Mar 17, 2026
69766c3
Parallelized post BLAST specificity check by chromsome
sfilges Mar 17, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 23 additions & 1 deletion src/plexus/blast/blast_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,23 @@ def _check_blast_tools() -> None:


# https://github.com/JasonAHendry/multiply/blob/master/src/multiply/blast/runner.py
_BLAST_DTYPES = {
"pident": "float32",
"length": "int32",
"mismatch": "int32",
"gapopen": "int32",
"qstart": "int32",
"qend": "int32",
"sstart": "int32", # human genome coords max ~250M, fits int32 (max 2.1B)
"send": "int32",
"evalue": "float32",
"bitscore": "float32",
"qlen": "int16",
}

_BLAST_CATEGORICAL_COLS = ("qseqid", "sseqid", "sstrand")


class BlastRunner:
BLAST_COLS = "qseqid sseqid pident length mismatch gapopen qstart qend sstart send evalue bitscore sstrand qlen"

Expand Down Expand Up @@ -205,8 +222,13 @@ def _load_as_dataframe(self):

# Load as a dataframe
self.blast_df = pd.read_csv(
self.output_table, sep="\t", names=self.BLAST_COLS.split(" ")
self.output_table,
sep="\t",
names=self.BLAST_COLS.split(" "),
dtype=_BLAST_DTYPES,
)
for col in _BLAST_CATEGORICAL_COLS:
self.blast_df[col] = self.blast_df[col].astype("category")

def get_dataframe(self):
"""
Expand Down
186 changes: 167 additions & 19 deletions src/plexus/blast/offtarget_finder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,39 +146,187 @@ def find_amplicons(self, max_size_bp=6000):
"R_target": r_targets,
"F_primer": np.full(n_matches, f_qseqid),
"R_primer": matched_r_qseqids,
"F_start": np.full(n_matches, f_start, dtype=int),
"R_start": matched_r_starts,
"product_bp": matched_r_starts - f_start + 1,
"F_start": np.full(n_matches, f_start, dtype=np.int32),
"R_start": matched_r_starts.astype(np.int32),
"product_bp": (matched_r_starts - f_start + 1).astype(np.int32),
"F_pident": np.full(
n_matches, f_pident[i] if f_pident is not None else None
n_matches,
f_pident[i] if f_pident is not None else np.nan,
dtype=np.float32,
),
"R_pident": (
r_pident[lo:hi].astype(np.float32)
if r_pident is not None
else np.full(n_matches, np.nan, dtype=np.float32)
),
"R_pident": r_pident[lo:hi]
if r_pident is not None
else np.full(n_matches, None),
"F_mismatch": np.full(
n_matches, f_mismatch[i] if f_mismatch is not None else None
n_matches,
f_mismatch[i] if f_mismatch is not None else -1,
dtype=np.int32,
),
"R_mismatch": (
r_mismatch[lo:hi].astype(np.int32)
if r_mismatch is not None
else np.full(n_matches, -1, dtype=np.int32)
),
"R_mismatch": r_mismatch[lo:hi]
if r_mismatch is not None
else np.full(n_matches, None),
"F_align_len": np.full(
n_matches, f_length[i] if f_length is not None else None
n_matches,
f_length[i] if f_length is not None else -1,
dtype=np.int32,
),
"R_align_len": (
r_length[lo:hi].astype(np.int32)
if r_length is not None
else np.full(n_matches, -1, dtype=np.int32)
),
"R_align_len": r_length[lo:hi]
if r_length is not None
else np.full(n_matches, None),
"F_evalue": np.full(
n_matches, f_evalue[i] if f_evalue is not None else None
n_matches,
f_evalue[i] if f_evalue is not None else np.nan,
dtype=np.float32,
),
"R_evalue": (
r_evalue[lo:hi].astype(np.float32)
if r_evalue is not None
else np.full(n_matches, np.nan, dtype=np.float32)
),
"R_evalue": r_evalue[lo:hi]
if r_evalue is not None
else np.full(n_matches, None),
}
result_chunks.append(pd.DataFrame(chunk, columns=amplicon_columns))

# Store
if result_chunks:
self.amplicon_df = pd.concat(result_chunks, ignore_index=True)
del result_chunks
for col in ("chrom", "F_target", "R_target", "F_primer", "R_primer"):
self.amplicon_df[col] = self.amplicon_df[col].astype("category")
else:
self.amplicon_df = pd.DataFrame(columns=amplicon_columns)

def find_amplicons_by_chrom(self, max_size_bp=6000):
"""Yield (chrom, amplicon_df) tuples, one per chromosome.

Same logic as find_amplicons() but yields per-chromosome results
instead of accumulating into self.amplicon_df. This bounds peak
memory to the largest single chromosome's amplicon set.
"""
amplicon_columns = [
"chrom",
"F_target",
"R_target",
"F_primer",
"R_primer",
"F_start",
"R_start",
"product_bp",
"F_pident",
"R_pident",
"F_mismatch",
"R_mismatch",
"F_align_len",
"R_align_len",
"F_evalue",
"R_evalue",
]
target_map = self.target_map

for chrom, chrom_df in self.bound_df.groupby("sseqid"):
fwd = chrom_df[chrom_df["sstrand"] == "plus"]
rev = chrom_df[chrom_df["sstrand"] == "minus"]

if fwd.empty or rev.empty:
continue

# Sort reverse hits by sstart for searchsorted
rev = rev.sort_values("sstart")
r_starts = rev["sstart"].values
r_qseqids = rev["qseqid"].values
r_pident = self._get_col_or_none(rev, "pident")
r_mismatch = self._get_col_or_none(rev, "mismatch")
r_length = self._get_col_or_none(rev, "length")
r_evalue = self._get_col_or_none(rev, "evalue")

f_starts = fwd["sstart"].values
f_qseqids = fwd["qseqid"].values
f_pident = self._get_col_or_none(fwd, "pident")
f_mismatch = self._get_col_or_none(fwd, "mismatch")
f_length = self._get_col_or_none(fwd, "length")
f_evalue = self._get_col_or_none(fwd, "evalue")

result_chunks = []

for i in range(len(f_starts)):
f_start = f_starts[i]
lo = np.searchsorted(r_starts, f_start, side="right")
hi = np.searchsorted(r_starts, f_start + max_size_bp, side="left")

if lo >= hi:
continue

n_matches = hi - lo
f_qseqid = f_qseqids[i]
f_target = target_map.get(f_qseqid, f_qseqid.split("_")[0])

matched_r_starts = r_starts[lo:hi]
matched_r_qseqids = r_qseqids[lo:hi]

r_targets = [
target_map.get(rq, rq.split("_")[0]) for rq in matched_r_qseqids
]

chunk = {
"chrom": np.full(n_matches, chrom),
"F_target": np.full(n_matches, f_target),
"R_target": r_targets,
"F_primer": np.full(n_matches, f_qseqid),
"R_primer": matched_r_qseqids,
"F_start": np.full(n_matches, f_start, dtype=np.int32),
"R_start": matched_r_starts.astype(np.int32),
"product_bp": (matched_r_starts - f_start + 1).astype(np.int32),
"F_pident": np.full(
n_matches,
f_pident[i] if f_pident is not None else np.nan,
dtype=np.float32,
),
"R_pident": (
r_pident[lo:hi].astype(np.float32)
if r_pident is not None
else np.full(n_matches, np.nan, dtype=np.float32)
),
"F_mismatch": np.full(
n_matches,
f_mismatch[i] if f_mismatch is not None else -1,
dtype=np.int32,
),
"R_mismatch": (
r_mismatch[lo:hi].astype(np.int32)
if r_mismatch is not None
else np.full(n_matches, -1, dtype=np.int32)
),
"F_align_len": np.full(
n_matches,
f_length[i] if f_length is not None else -1,
dtype=np.int32,
),
"R_align_len": (
r_length[lo:hi].astype(np.int32)
if r_length is not None
else np.full(n_matches, -1, dtype=np.int32)
),
"F_evalue": np.full(
n_matches,
f_evalue[i] if f_evalue is not None else np.nan,
dtype=np.float32,
),
"R_evalue": (
r_evalue[lo:hi].astype(np.float32)
if r_evalue is not None
else np.full(n_matches, np.nan, dtype=np.float32)
),
}
result_chunks.append(pd.DataFrame(chunk, columns=amplicon_columns))

if result_chunks:
chrom_amplicon_df = pd.concat(result_chunks, ignore_index=True)
del result_chunks
for col in ("chrom", "F_target", "R_target", "F_primer", "R_primer"):
chrom_amplicon_df[col] = chrom_amplicon_df[col].astype("category")
yield chrom, chrom_amplicon_df
Loading
Loading