phu 0.4.2__tar.gz → 0.4.4__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: phu
3
- Version: 0.4.2
3
+ Version: 0.4.4
4
4
  Summary: Phage bioinformatics utilities (seqclust runner and friends).
5
5
  Author-email: Camilo García-Botero <ca.garcia2@uniandes.edu.co>
6
6
  License: MIT
@@ -8,7 +8,9 @@ Requires-Python: >=3.10
8
8
  Description-Content-Type: text/markdown
9
9
  License-File: LICENSE
10
10
  Requires-Dist: pandas>=2.3.2
11
+ Requires-Dist: pyhmmer>=0.11.1
11
12
  Requires-Dist: pyrodigal>=3.6.3.post1
13
+ Requires-Dist: pyrodigal-gv>=0.3.2
12
14
  Requires-Dist: typer>=0.17.3
13
15
  Dynamic: license-file
14
16
 
@@ -1,6 +1,6 @@
1
1
  [project]
2
2
  name = "phu"
3
- version = "0.4.2"
3
+ version = "0.4.4"
4
4
  description = "Phage bioinformatics utilities (seqclust runner and friends)."
5
5
  readme = "README.md"
6
6
  license = { text = "MIT" }
@@ -10,7 +10,9 @@ authors = [
10
10
  requires-python = ">=3.10"
11
11
  dependencies = [
12
12
  "pandas>=2.3.2",
13
+ "pyhmmer>=0.11.1",
13
14
  "pyrodigal>=3.6.3.post1",
15
+ "pyrodigal-gv>=0.3.2",
14
16
  "typer>=0.17.3",
15
17
  ]
16
18
 
@@ -185,6 +185,9 @@ def screen(
185
185
  False, "--save-target-proteins/--no-save-target-proteins",
186
186
  help="Save matched proteins per HMM model in target_proteins/ subfolder"
187
187
  ),
188
+ save_target_hmms: bool = typer.Option(
189
+ False, "--save-target-hmms/--no-save-target-hmms", help="Save HMMs built from target proteins in target_hmms/ subfolder"
190
+ ),
188
191
  hmm_mode: str = typer.Option(
189
192
  "pure", "--hmm-mode", help="HMM file type: 'pure' (one model per file) or 'mixed' (pressed/concatenated HMMs)"
190
193
  )
@@ -238,6 +241,7 @@ def screen(
238
241
  combine_mode=combine_mode,
239
242
  min_hmm_hits=min_hmm_hits,
240
243
  save_target_proteins=save_target_proteins,
244
+ save_target_hmms=save_target_hmms,
241
245
  hmm_mode=hmm_mode, # New parameter
242
246
  )
243
247
 
@@ -10,13 +10,19 @@ from dataclasses import dataclass, field
10
10
  from pathlib import Path
11
11
  from typing import Dict, Iterable, List, Optional, Tuple
12
12
  from multiprocessing.pool import ThreadPool
13
+ from collections import defaultdict
13
14
 
14
15
  import typer
15
16
  from pyrodigal import GeneFinder # pyrodigal>=3
17
+ from pyrodigal_gv import ViralGeneFinder # New import for viral gene prediction
18
+
19
+ import pyhmmer
20
+ import pyhmmer.plan7
21
+ import pyhmmer.easel
16
22
 
17
23
  from ._exec import run, _executable, CmdNotFound
18
24
 
19
- app = typer.Typer(help="Screen contigs for a protein family using HMMER on predicted CDS.")
25
+ app = typer.Typer(help="Screen contigs for a protein family using pyHMMER on predicted CDS.")
20
26
 
21
27
 
22
28
  # ---------- Utilities ----------
@@ -42,6 +48,7 @@ class ScreenConfig:
42
48
  combine_mode: str = "any" # New: how to combine hits from multiple HMMs
43
49
  min_hmm_hits: int = 1 # New: minimum number of HMMs that must hit a contig
44
50
  save_target_proteins: bool = False # New: save matched proteins per HMM
51
+ save_target_hmms: bool = False # New: build and save HMMs from target proteins
45
52
  hmm_mode: str = "pure" # New: "pure" for single models, "mixed" for pressed/concatenated HMMs
46
53
 
47
54
  def __post_init__(self):
@@ -58,7 +65,9 @@ class ScreenConfig:
58
65
  raise ValueError("combine_mode must be 'any', 'all', or 'threshold'")
59
66
  if self.hmm_mode not in {"pure", "mixed"}:
60
67
  raise ValueError("hmm_mode must be 'pure' or 'mixed'")
61
-
68
+ if self.save_target_hmms and not self.save_target_proteins:
69
+ raise ValueError("save_target_hmms requires save_target_proteins to be True")
70
+
62
71
  def plan(self) -> "ScreenPlan":
63
72
  """Create execution plan from configuration."""
64
73
  if self.mode not in {"meta", "single"}:
@@ -90,6 +99,7 @@ class ScreenConfig:
90
99
  combine_mode=self.combine_mode,
91
100
  min_hmm_hits=self.min_hmm_hits,
92
101
  save_target_proteins=self.save_target_proteins,
102
+ save_target_hmms=self.save_target_hmms,
93
103
  hmm_mode=self.hmm_mode,
94
104
  proteins_fa=self.outdir / "proteins.faa",
95
105
  domtbl_paths=domtbl_paths,
@@ -118,19 +128,20 @@ class ScreenPlan:
118
128
  combine_mode: str
119
129
  min_hmm_hits: int
120
130
  save_target_proteins: bool # New
131
+ save_target_hmms: bool # New
121
132
  hmm_mode: str # New
122
133
  proteins_fa: Path
123
134
  domtbl_paths: Dict[str, Path] # Changed from domtbl: Path
124
135
  kept_ids: Path
125
136
  out_contigs: Path
126
137
 
127
- def _binaries() -> tuple[str, str]:
138
+ def _binaries() -> str:
128
139
  """
129
140
  Discover required binaries for screening.
141
+ Only seqkit is needed since we use pyHMMER instead of HMMER binary.
130
142
  """
131
- hmmer = _executable(["hmmsearch"])
132
143
  seqkit = _executable(["seqkit"])
133
- return hmmer, seqkit
144
+ return seqkit
134
145
 
135
146
  def _read_fasta(fp: Path) -> Iterable[Tuple[str, str]]:
136
147
  """Tiny FASTA reader (header up to first whitespace is the id)."""
@@ -155,57 +166,6 @@ class Hit:
155
166
  bitscore: float
156
167
  evalue: float
157
168
 
158
- def _parse_domtblout(domtbl_path: Path, hmm_file_name: str, hmm_mode: str) -> Iterable[Hit]:
159
- """
160
- Parse HMMER --domtblout (hmmsearch). Returns domain-level hits.
161
- Format spec: https://hmmer-web-docs.readthedocs.io/en/latest/output-files/domtab.html
162
-
163
- Args:
164
- hmm_mode: "pure" for single-model HMMs, "mixed" for concatenated HMMs
165
-
166
- Filtering logic:
167
- - Individual hits are filtered by bitscore/evalue thresholds regardless of hmm_mode
168
- - Model counting for combine_mode depends on hmm_mode:
169
- * "pure": each file = 1 model, use filename as identifier
170
- * "mixed": each target_name = 1 model, use actual HMM name as identifier
171
- """
172
- with domtbl_path.open() as f:
173
- for line in f:
174
- if not line or line.startswith("#"):
175
- continue
176
- cols = re.split(r"\s+", line.strip())
177
- try:
178
- target_name = cols[0] # our protein sequence ID
179
- query_name = cols[3] # HMM model name
180
- i_evalue = float(cols[12]) # independent E-value
181
- bits = float(cols[13])
182
- except Exception:
183
- # Fallback for slight index shifts (older HMMER versions)
184
- try:
185
- i_evalue = float(cols[11])
186
- bits = float(cols[12])
187
- except Exception:
188
- continue
189
-
190
- # Extract contig ID from protein sequence ID (query_name is the protein ID)
191
- prot_id = target_name # This should be like "contig123|gene1"
192
- if "|" in prot_id:
193
- contig_id = prot_id.split("|", 1)[0]
194
- else:
195
- contig_id = prot_id
196
-
197
- # Handle model naming based on HMM mode
198
- if hmm_mode == "pure":
199
- # For pure HMMs, use filename as the model identifier for consistency
200
- # This ensures each file is treated as a separate "model" for combine logic
201
- model_id = hmm_file_name
202
- else: # mixed
203
- # For mixed/pressed HMMs, use the actual HMM model name (query_name)
204
- # This allows proper identification of individual models within the file
205
- model_id = query_name
206
-
207
- yield Hit(contig=contig_id, prot_id=prot_id, model=model_id, bitscore=bits, evalue=i_evalue)
208
-
209
169
 
210
170
  # ---------- Core pipeline ----------
211
171
 
@@ -225,7 +185,7 @@ def _predict_proteins_pyrodigal(
225
185
  Uses ThreadPool for parallel processing of contigs when threads > 1.
226
186
  """
227
187
  # Initialize GeneFinder according to the API
228
- gf = GeneFinder(meta=(mode == "meta"), min_gene=min_len)
188
+ gf = ViralGeneFinder(meta=(mode == "meta"), min_gene=min_len)
229
189
 
230
190
  # Read all contigs first
231
191
  contigs = list(_read_fasta(contigs_fa))
@@ -270,58 +230,84 @@ def _predict_proteins_pyrodigal(
270
230
  return n_prot
271
231
 
272
232
  def _hmmsearch(
273
- hmm: Path,
233
+ hmm_paths: List[Path],
274
234
  proteins_fa: Path,
275
- domtbl_path: Path,
235
+ domtbl_paths: Dict[str, Path],
276
236
  threads: int = 1,
277
- hmmer_bin: str = "hmmsearch",
278
- extra_args: Optional[List[str]] = None,
279
- ) -> None:
280
- """Run hmmsearch with proper HMMER command structure."""
281
- # Validate inputs before running
282
- if not hmm.exists():
283
- raise FileNotFoundError(f"HMM file not found: {hmm}")
284
- if not proteins_fa.exists():
285
- raise FileNotFoundError(f"Protein FASTA file not found: {proteins_fa}")
286
- if proteins_fa.stat().st_size == 0:
287
- raise ValueError(f"Protein FASTA file is empty: {proteins_fa}")
288
-
289
- # Build command according to HMMER documentation:
290
- # hmmsearch [options] <hmmfile> <seqfile>
291
- cmd = [hmmer_bin]
292
-
293
- # Add all options first
294
- # Note: --cpu <n> specifies number of worker threads
295
- # HMMER will spawn <n>+1 total threads (workers + master)
296
- cmd.extend(["--cpu", str(threads)])
297
- cmd.extend(["--domtblout", str(domtbl_path)])
298
-
299
- if extra_args:
300
- cmd.extend(extra_args)
301
-
302
- # Add positional arguments: HMM file first, then sequence file
303
- cmd.append(str(hmm))
304
- cmd.append(str(proteins_fa))
305
-
306
- # Use subprocess directly for better error handling
307
- try:
308
- result = subprocess.run(
309
- cmd,
310
- capture_output=True,
311
- text=True,
312
- check=True
313
- )
314
- # Log successful completion for debugging
315
- if result.returncode == 0:
316
- print(f" hmmsearch completed successfully using {threads} worker threads")
317
- except subprocess.CalledProcessError as e:
318
- print(f"hmmsearch failed with exit code {e.returncode}")
319
- print(f"Command: {' '.join(cmd)}")
320
- if e.stdout:
321
- print(f"STDOUT:\n{e.stdout}")
322
- if e.stderr:
323
- print(f"STDERR:\n{e.stderr}")
324
- raise RuntimeError(f"hmmsearch failed: {e.stderr}") from e
237
+ hmm_mode: str = "pure",
238
+ keep_domtbl: bool = True,
239
+ ) -> Iterable[Hit]:
240
+ """
241
+ Run pyhmmer.hmmsearch on loaded HMMs and proteins.
242
+ Returns hits directly as Hit objects and optionally writes domtbl files.
243
+ """
244
+ # Load all HMMs into memory
245
+ hmms = []
246
+ hmm_names = []
247
+ for hmm_path in hmm_paths:
248
+ with pyhmmer.plan7.HMMFile(hmm_path) as hmm_file:
249
+ for hmm in hmm_file:
250
+ hmms.append(hmm)
251
+ hmm_names.append(hmm_path.stem) # Use filename for pure mode
252
+
253
+ # Load proteins into memory
254
+ with pyhmmer.easel.SequenceFile(proteins_fa, digital=True) as seq_file:
255
+ proteins = seq_file.read_block()
256
+
257
+ # Run hmmsearch with pyHMMER
258
+ hits_list = list(pyhmmer.hmmsearch(hmms, proteins, cpus=threads, bit_cutoffs=None))
259
+
260
+ # Write domtbl files if requested
261
+ if keep_domtbl:
262
+ for i, top_hits in enumerate(hits_list):
263
+ # Map back to original HMM file for domtbl naming
264
+ hmm_name = hmm_names[i] if i < len(hmm_names) else f"hmm_{i}"
265
+ if hmm_name in domtbl_paths:
266
+ domtbl_path = domtbl_paths[hmm_name]
267
+ with domtbl_path.open("wb") as f:
268
+ top_hits.write(f, format="domains")
269
+
270
+ # Process hits and yield Hit objects
271
+ for i, top_hits in enumerate(hits_list):
272
+ model_name = top_hits.query.name.decode()
273
+
274
+ # Determine model identifier based on hmm_mode
275
+ if hmm_mode == "pure":
276
+ # Use filename as model ID for pure mode
277
+ model_id = hmm_names[i] if i < len(hmm_names) else model_name
278
+ else:
279
+ # Use actual HMM name for mixed mode
280
+ model_id = model_name
281
+
282
+ for hit in top_hits:
283
+ if hit.included: # pyHMMER's inclusion check
284
+ prot_id = hit.name.decode()
285
+
286
+ # Extract contig from prot_id with robust handling of multiple "|" characters
287
+ # Expected format: "contig_name|gene<idx>" where contig_name may contain "|"
288
+ # Use regex to find the last "|gene" pattern
289
+ import re
290
+ gene_pattern = r'\|gene\d+$'
291
+ match = re.search(gene_pattern, prot_id)
292
+ if match:
293
+ # Split at the position where "|gene" starts
294
+ contig = prot_id[:match.start()]
295
+ else:
296
+ # Fallback: if no "|gene" pattern found, try simple split
297
+ if "|" in prot_id:
298
+ # Take everything before the last "|" as contig ID
299
+ contig = prot_id.rsplit("|", 1)[0]
300
+ else:
301
+ # No "|" found, use entire protein ID as contig ID
302
+ contig = prot_id
303
+
304
+ yield Hit(
305
+ contig=contig,
306
+ prot_id=prot_id,
307
+ model=model_id,
308
+ bitscore=hit.score,
309
+ evalue=hit.evalue
310
+ )
325
311
 
326
312
  def _choose_best_contigs(
327
313
  hits: Iterable[Hit],
@@ -336,33 +322,11 @@ def _choose_best_contigs(
336
322
  Filter by thresholds, then pick top N hits per contig by bitscore.
337
323
  For multiple HMMs, apply combination logic.
338
324
 
339
- Filtering criteria with hmm_mode:
340
-
341
- 1. Individual hit filtering (same for both modes):
342
- - min_bitscore: minimum score threshold per hit
343
- - max_evalue: maximum E-value threshold per hit
344
-
345
- 2. Model counting for combine logic (depends on hmm_mode):
346
- - "pure" mode: each HMM file = 1 model
347
- * "any": contig needs hits from ≥1 files
348
- * "all": contig needs hits from ALL files
349
- * "threshold": contig needs hits from ≥min_hmm_hits files
350
-
351
- - "mixed" mode: each unique target name = 1 model
352
- * "any": contig needs hits from ≥1 unique models
353
- * "all": contig needs hits from ALL unique models found across all files
354
- * "threshold": contig needs hits from ≥min_hmm_hits unique models
355
-
356
- 3. Hit selection per contig:
357
- - For "any" and "threshold": top_per_contig best hits overall
358
- - For "all": exactly 1 best hit per model (ensures balanced output)
359
-
360
325
  Returns (kept_hits, list_of_contig_ids).
361
326
  """
362
- from collections import defaultdict
363
327
  per_contig: Dict[str, List[Hit]] = defaultdict(list)
364
328
 
365
- # Filter hits by thresholds (same for both hmm modes)
329
+ # Filter hits by thresholds
366
330
  for h in hits:
367
331
  if min_bitscore is not None and h.bitscore < min_bitscore:
368
332
  continue
@@ -370,17 +334,12 @@ def _choose_best_contigs(
370
334
  continue
371
335
  per_contig[h.contig].append(h)
372
336
 
373
- # Apply combination logic (behavior depends on hmm_mode through model_id assignment)
337
+ # Apply combination logic
374
338
  kept: List[Hit] = []
375
339
  kept_contigs: List[str] = []
376
340
 
377
341
  for contig, contig_hits in per_contig.items():
378
342
  if combine_mode == "any":
379
- # Keep contigs that have hits from any model.
380
- # Instead of keeping only the single best hit across all models, keep
381
- # the best hit per model/file. This recovers one (or up to
382
- # top_per_contig) protein per model for contigs that matched
383
- # multiple models.
384
343
  if contig_hits:
385
344
  hits_per_model = defaultdict(list)
386
345
  for hit in contig_hits:
@@ -395,17 +354,15 @@ def _choose_best_contigs(
395
354
  elif combine_mode == "all":
396
355
  # Keep contigs that have hits from ALL models
397
356
  model_names = set(hit.model for hit in contig_hits)
398
- if len(model_names) == total_hmm_models: # Must have hits from ALL models
399
- # For "all" mode, ensure we get exactly one hit per model per contig
357
+ if len(model_names) == total_hmm_models:
400
358
  hits_per_model = defaultdict(list)
401
359
  for hit in contig_hits:
402
- model_name = hit.model
403
- hits_per_model[model_name].append(hit)
360
+ hits_per_model[hit.model].append(hit)
404
361
 
405
- # Take the best hit for each model (ensures balanced protein counts)
406
- for model_name, model_hits in hits_per_model.items():
362
+ # Take the best hit for each model
363
+ for model_hits in hits_per_model.values():
407
364
  model_hits.sort(key=lambda x: (x.bitscore, -x.evalue), reverse=True)
408
- kept.extend(model_hits[:1]) # Take only the best hit per model
365
+ kept.extend(model_hits[:1])
409
366
 
410
367
  kept_contigs.append(contig)
411
368
 
@@ -414,8 +371,7 @@ def _choose_best_contigs(
414
371
  model_names = set(hit.model for hit in contig_hits)
415
372
  if len(model_names) >= min_hmm_hits:
416
373
  contig_hits.sort(key=lambda x: (x.bitscore, -x.evalue), reverse=True)
417
- slice_ = contig_hits[:max(1, top_per_contig)]
418
- kept.extend(slice_)
374
+ kept.extend(contig_hits[:max(1, top_per_contig)])
419
375
  kept_contigs.append(contig)
420
376
 
421
377
  return kept, kept_contigs
@@ -450,15 +406,7 @@ def _extract_target_proteins(
450
406
  ) -> None:
451
407
  """
452
408
  Extract matched proteins per HMM model from the final kept contigs only.
453
- This ensures the proteins match those from contigs in screened_contigs.fasta
454
- and respects the combine_mode filtering.
455
-
456
- Protein extraction behavior:
457
- - "pure" mode: proteins grouped by HMM filename (one group per file)
458
- - "mixed" mode: proteins grouped by actual model names (multiple groups per file possible)
459
409
  """
460
- from collections import defaultdict
461
-
462
410
  # Create a set of kept contig IDs for fast lookup
463
411
  kept_contig_set = set(kept_contig_ids)
464
412
 
@@ -466,10 +414,8 @@ def _extract_target_proteins(
466
414
  proteins_per_model: Dict[str, List[str]] = defaultdict(list)
467
415
 
468
416
  for hit in kept_hits:
469
- # Only include proteins from contigs that actually made it to the final output
470
417
  if hit.contig in kept_contig_set:
471
- model_id = hit.model
472
- proteins_per_model[model_id].append(hit.prot_id)
418
+ proteins_per_model[hit.model].append(hit.prot_id)
473
419
 
474
420
  # Create target_proteins directory
475
421
  target_proteins_dir = outdir / "target_proteins"
@@ -482,7 +428,6 @@ def _extract_target_proteins(
482
428
  output_path = target_proteins_dir / f"{safe_model_name}_proteins.mfa"
483
429
 
484
430
  if not protein_ids:
485
- # Write empty file
486
431
  output_path.write_text("")
487
432
  continue
488
433
 
@@ -511,14 +456,112 @@ def _extract_target_proteins(
511
456
  # Cleanup temporary file
512
457
  tmp_ids_file.unlink()
513
458
 
459
+ def _build_target_hmms(
460
+ target_proteins_dir: Path,
461
+ outdir: Path,
462
+ threads: int = 1,
463
+ ) -> None:
464
+ """
465
+ Build HMM models from target protein sequences using pyHMMER.
466
+ Creates one HMM file per model from the corresponding protein FASTA files.
467
+
468
+ For single sequences, builds HMM directly using builder.build().
469
+ For multiple sequences, aligns them by padding to the same length before MSA creation.
470
+ """
471
+ # Create target_hmms directory
472
+ target_hmms_dir = outdir / "target_hmms"
473
+ target_hmms_dir.mkdir(parents=True, exist_ok=True)
474
+
475
+ # Find all protein FASTA files in target_proteins directory
476
+ protein_files = list(target_proteins_dir.glob("*_proteins.mfa"))
477
+
478
+ if not protein_files:
479
+ print(" No target protein files found for HMM building")
480
+ return
481
+
482
+ # Initialize HMM builder and background
483
+ alphabet = pyhmmer.easel.Alphabet.amino()
484
+ builder = pyhmmer.plan7.Builder(alphabet)
485
+ background = pyhmmer.plan7.Background(alphabet)
486
+
487
+ print(f" Building HMMs for {len(protein_files)} protein sets...")
488
+
489
+ for protein_file in protein_files:
490
+ model_name = protein_file.stem.replace("_proteins", "")
491
+ hmm_output_path = target_hmms_dir / f"{model_name}.hmm"
492
+
493
+ try:
494
+ # Read protein sequences as text first
495
+ sequences = []
496
+ with open(protein_file, 'r') as f:
497
+ seq_id, seq_chunks = None, []
498
+ for line in f:
499
+ line = line.strip()
500
+ if line.startswith('>'):
501
+ if seq_id is not None and seq_chunks:
502
+ sequence = ''.join(seq_chunks)
503
+ if sequence: # Only add non-empty sequences
504
+ text_seq = pyhmmer.easel.TextSequence(name=seq_id.encode(), sequence=sequence)
505
+ sequences.append(text_seq)
506
+ seq_id = line[1:].split()[0]
507
+ seq_chunks = []
508
+ else:
509
+ seq_chunks.append(line)
510
+
511
+ # Don't forget the last sequence
512
+ if seq_id is not None and seq_chunks:
513
+ sequence = ''.join(seq_chunks)
514
+ if sequence:
515
+ text_seq = pyhmmer.easel.TextSequence(name=seq_id.encode(), sequence=sequence)
516
+ sequences.append(text_seq)
517
+
518
+ if len(sequences) == 0:
519
+ print(f" Skipping {model_name}: no valid sequences found")
520
+ continue
521
+ elif len(sequences) == 1:
522
+ # For single sequence, use builder.build() method
523
+ digital_seq = sequences[0].digitize(alphabet)
524
+ hmm = builder.build(digital_seq, background)
525
+ hmm.name = model_name.encode()
526
+ print(f" Built HMM from 1 sequence: {model_name}")
527
+ else:
528
+ # For multiple sequences, align by padding to same length
529
+ max_len = max(len(seq.sequence) for seq in sequences)
530
+
531
+ # Pad sequences to same length with gaps
532
+ aligned_sequences = []
533
+ for seq in sequences:
534
+ padded_seq = seq.sequence + '-' * (max_len - len(seq.sequence))
535
+ aligned_seq = pyhmmer.easel.TextSequence(name=seq.name, sequence=padded_seq)
536
+ aligned_sequences.append(aligned_seq)
537
+
538
+ # Create MSA from aligned sequences
539
+ text_msa = pyhmmer.easel.TextMSA(name=model_name.encode(), sequences=aligned_sequences)
540
+ digital_msa = text_msa.digitize(alphabet)
541
+ hmm, _, _ = builder.build_msa(digital_msa, background)
542
+ print(f" Built HMM from {len(sequences)} aligned sequences: {model_name}")
543
+
544
+ # Ensure HMM has proper name
545
+ if not hmm.name:
546
+ hmm.name = model_name.encode()
547
+
548
+ # Write HMM to file
549
+ with hmm_output_path.open("wb") as f:
550
+ hmm.write(f)
551
+
552
+ except Exception as e:
553
+ print(f" Warning: Failed to build HMM for {model_name}: {e}")
554
+ import traceback
555
+ traceback.print_exc()
556
+ continue
514
557
 
515
558
  def _screen(cfg: ScreenConfig) -> ScreenPlan:
516
559
  """
517
- Screen contigs for protein families using HMMER.
560
+ Screen contigs for protein families using pyHMMER.
518
561
 
519
562
  Main workflow:
520
563
  1. Predict proteins with pyrodigal
521
- 2. Search proteins against each HMM with hmmsearch
564
+ 2. Search proteins against HMMs with pyhmmer.hmmsearch
522
565
  3. Parse results, combine, and select best hits per contig
523
566
  4. Extract screened contigs
524
567
  5. Optionally extract matched proteins per HMM model from screened contigs only
@@ -531,10 +574,10 @@ def _screen(cfg: ScreenConfig) -> ScreenPlan:
531
574
  if not hmm.exists():
532
575
  raise FileNotFoundError(f"HMM file not found: {hmm}")
533
576
 
534
- # Discover binaries
535
- hmmer_bin, seqkit_bin = _binaries()
577
+ # Discover binaries (only seqkit needed)
578
+ seqkit_bin = _binaries()
536
579
  plan = cfg.plan()
537
- plan.hmmer_bin = hmmer_bin
580
+ plan.hmmer_bin = "" # Not used with pyHMMER
538
581
  plan.seqkit_bin = seqkit_bin
539
582
 
540
583
  # Create output directory
@@ -547,7 +590,7 @@ def _screen(cfg: ScreenConfig) -> ScreenPlan:
547
590
  mode=plan.mode,
548
591
  min_len=plan.min_gene_len,
549
592
  translation_table=plan.translation_table,
550
- threads=plan.threads, # Added threads parameter
593
+ threads=plan.threads,
551
594
  )
552
595
  print(f" Proteins predicted: {n_prot}")
553
596
 
@@ -559,33 +602,19 @@ def _screen(cfg: ScreenConfig) -> ScreenPlan:
559
602
  plan.proteins_fa.unlink()
560
603
  return plan
561
604
 
562
- print(f"Running hmmsearch for {len(plan.hmms)} HMM file(s) (mode: {plan.hmm_mode})…")
563
- all_hits = []
564
- unique_model_ids = set()
605
+ print(f"Running pyhmmer.hmmsearch for {len(plan.hmms)} HMM file(s) (mode: {plan.hmm_mode})…")
565
606
 
566
- for hmm in plan.hmms:
567
- hmm_file_name = hmm.stem
568
- domtbl_path = plan.domtbl_paths[hmm_file_name]
569
- print(f" Searching with {hmm.name}...")
570
-
571
- _hmmsearch(
572
- hmm=hmm,
573
- proteins_fa=plan.proteins_fa,
574
- domtbl_path=domtbl_path,
575
- threads=plan.threads,
576
- hmmer_bin=plan.hmmer_bin,
577
- extra_args=["--noali"],
578
- )
579
-
580
- hits = list(_parse_domtblout(domtbl_path, hmm_file_name, plan.hmm_mode))
581
- all_hits.extend(hits)
582
-
583
- # Collect unique model identifiers
584
- for hit in hits:
585
- unique_model_ids.add(hit.model)
586
-
587
- print(f" Found {len(hits)} hits")
607
+ # Use pyHMMER for all searches
608
+ all_hits = list(_hmmsearch(
609
+ hmm_paths=plan.hmms,
610
+ proteins_fa=plan.proteins_fa,
611
+ domtbl_paths=plan.domtbl_paths,
612
+ threads=plan.threads,
613
+ hmm_mode=plan.hmm_mode,
614
+ keep_domtbl=plan.keep_domtbl,
615
+ ))
588
616
 
617
+ unique_model_ids = set(hit.model for hit in all_hits)
589
618
  if plan.hmm_mode == "pure":
590
619
  total_models = len(plan.hmms) # Each file is one model
591
620
  print(f" Pure HMM mode: {total_models} models from {len(plan.hmms)} files")
@@ -593,6 +622,8 @@ def _screen(cfg: ScreenConfig) -> ScreenPlan:
593
622
  total_models = len(unique_model_ids) # Count actual models found
594
623
  print(f" Mixed HMM mode: {total_models} unique models found from {len(plan.hmms)} files")
595
624
 
625
+ print(f" Found {len(all_hits)} hits")
626
+
596
627
  print(f"Parsing results and selecting best hits per contig (combine_mode: {plan.combine_mode})…")
597
628
  kept_hits, contig_ids = _choose_best_contigs(
598
629
  all_hits,
@@ -613,10 +644,20 @@ def _screen(cfg: ScreenConfig) -> ScreenPlan:
613
644
  kept_hits,
614
645
  contig_ids,
615
646
  plan.proteins_fa,
616
- plan.outdir, # Pass outdir for proper path construction
647
+ plan.outdir,
617
648
  plan.hmm_mode,
618
649
  plan.seqkit_bin,
619
650
  )
651
+
652
+ # Build HMMs from target proteins if requested
653
+ if plan.save_target_hmms:
654
+ print("Building HMMs from target protein sequences…")
655
+ target_proteins_dir = plan.outdir / "target_proteins"
656
+ _build_target_hmms(
657
+ target_proteins_dir,
658
+ plan.outdir,
659
+ threads=plan.threads,
660
+ )
620
661
 
621
662
  print(f"Extracting {len(contig_ids)} contig(s) with seqkit…")
622
663
  _seqkit_extract(
@@ -640,6 +681,8 @@ def _screen(cfg: ScreenConfig) -> ScreenPlan:
640
681
  files_msg += f" and {len(plan.domtbl_paths)} domtblout files"
641
682
  if plan.save_target_proteins:
642
683
  files_msg += f" and target proteins in target_proteins/ folder"
684
+ if plan.save_target_hmms:
685
+ files_msg += f" and target HMMs in target_hmms/ folder"
643
686
  print(f"{files_msg}.")
644
687
 
645
688
  return plan
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: phu
3
- Version: 0.4.2
3
+ Version: 0.4.4
4
4
  Summary: Phage bioinformatics utilities (seqclust runner and friends).
5
5
  Author-email: Camilo García-Botero <ca.garcia2@uniandes.edu.co>
6
6
  License: MIT
@@ -8,7 +8,9 @@ Requires-Python: >=3.10
8
8
  Description-Content-Type: text/markdown
9
9
  License-File: LICENSE
10
10
  Requires-Dist: pandas>=2.3.2
11
+ Requires-Dist: pyhmmer>=0.11.1
11
12
  Requires-Dist: pyrodigal>=3.6.3.post1
13
+ Requires-Dist: pyrodigal-gv>=0.3.2
12
14
  Requires-Dist: typer>=0.17.3
13
15
  Dynamic: license-file
14
16
 
@@ -1,3 +1,5 @@
1
1
  pandas>=2.3.2
2
+ pyhmmer>=0.11.1
2
3
  pyrodigal>=3.6.3.post1
4
+ pyrodigal-gv>=0.3.2
3
5
  typer>=0.17.3
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes