From 14790d4c880248f844e2d300a752511d18eeff1a Mon Sep 17 00:00:00 2001
From: Stefan Bienert <stefan.bienert@unibas.ch>
Date: Fri, 12 Jan 2024 15:47:11 +0100
Subject: [PATCH] More fine grained software parameters
---
convert_to_modelcif.py | 126 +++++++++++++++++++++++++++++++----------
1 file changed, 96 insertions(+), 30 deletions(-)
diff --git a/convert_to_modelcif.py b/convert_to_modelcif.py
index 974d594..27b7894 100755
--- a/convert_to_modelcif.py
+++ b/convert_to_modelcif.py
@@ -644,24 +644,74 @@ def _cast_param(val):
def _get_software_with_parameters(sw_dict, other_dict):
"""Get software with versions and parameters."""
+
# ToDo: deal with `use_mmseqs=True`
# ToDo: should all args go to AlphaPulldown?
+ def _assemble_params(key, known_args, swwp):
+ for mthd in known_args[key]["method_type"]:
+ for tool in known_args[key]["sw"]:
+ if mthd not in swwp[tool]:
+ swwp[tool][mthd] = {}
+ swwp[tool][mthd][f"--{key}"] = _cast_param(val)
+
known_args = {
- "db_preset": ["AlphaFold"],
- "max_template_date": ["AlphaFold"],
- "model_preset": ["AlphaFold"],
- "num_multimer_predictions_per_model": ["AlphaFold"],
- "plddt_threshold": ["AlphaPulldown"],
- "hb_allowance": ["AlphaPulldown"],
- "threshold_clashes": ["AlphaPulldown"],
- "job_index": ["AlphaPulldown"],
- "use_mmseqs2": ["AlphaPulldown"],
- "skip_existing": ["AlphaPulldown"],
- "save_msa_files": ["AlphaPulldown"],
- "num_predictions_per_model": ["AlphaPulldown"],
- "benchmark": ["AlphaFold", "AlphaPulldown"],
- "use_precomputed_msas": ["AlphaFold", "AlphaPulldown"],
- "models_to_relax": ["AlphaFold", "AlphaPulldown"],
+ "db_preset": {"sw": ["AlphaFold"], "method_type": ["coevolution MSA"]},
+ "max_template_date": {
+ "sw": ["AlphaFold"],
+ "method_type": ["coevolution MSA"],
+ },
+ "model_preset": {
+ "sw": ["AlphaFold"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "num_multimer_predictions_per_model": {
+ "sw": ["AlphaFold"],
+ "method_type": ["modeling"],
+ },
+ "plddt_threshold": {
+ "sw": ["AlphaPulldown"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "hb_allowance": {
+ "sw": ["AlphaPulldown"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "threshold_clashes": {
+ "sw": ["AlphaPulldown"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "job_index": {
+ "sw": ["AlphaPulldown"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "use_mmseqs2": {
+ "sw": ["AlphaPulldown"],
+ "method_type": ["coevolution MSA"],
+ },
+ "skip_existing": {
+ "sw": ["AlphaPulldown"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "save_msa_files": {
+ "sw": ["AlphaPulldown"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "num_predictions_per_model": {
+ "sw": ["AlphaPulldown"],
+ "method_type": ["modeling"],
+ },
+ "benchmark": {
+ "sw": ["AlphaFold", "AlphaPulldown"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "use_precomputed_msas": {
+ "sw": ["AlphaFold", "AlphaPulldown"],
+ "method_type": ["coevolution MSA", "modeling"],
+ },
+ "models_to_relax": {
+ "sw": ["AlphaFold", "AlphaPulldown"],
+ "method_type": ["modeling"],
+ },
}
trans_args = {
"num_multimer_predictions_per_model": "num_predictions_per_model"
@@ -717,16 +767,10 @@ def _get_software_with_parameters(sw_dict, other_dict):
swwp = sw_dict # Software With Parameters
for key, val in other_dict.items():
if key in known_args:
- for tool in known_args[key]:
- if "params" not in swwp[tool]:
- swwp[tool]["params"] = {}
- swwp[tool]["params"][f"--{key}"] = _cast_param(val)
+ _assemble_params(key, known_args, swwp)
if key in trans_args:
key = trans_args[key]
- for tool in known_args[key]:
- if "params" not in swwp[tool]:
- swwp[tool]["params"] = {}
- swwp[tool]["params"][f"--{key}"] = _cast_param(val)
+ _assemble_params(key, known_args, swwp)
else:
if key not in ignored_args and re.match(re_args, key) is None:
logging.info(f"Found unknown key in 'other': {key}")
@@ -1066,28 +1110,50 @@ def _get_protocol_steps(modelcif_json):
for tool in sftwr:
if tool not in step["software_group"]:
step["software_group"].append(tool)
- if "params" in sftwr[tool]:
- step["parameter_group"].append(sftwr[tool]["params"])
+ if step["method_type"] in sftwr[tool]:
+ step["parameter_group"].append(
+ sftwr[tool][step["method_type"]]
+ )
else:
step["parameter_group"].append({})
+ else:
+ pos = step["software_group"].index(tool)
+ if step["method_type"] in sftwr[tool]:
+ params = sftwr[tool][step["method_type"]]
+ else:
+ params = {}
+ if step["parameter_group"][pos] != params:
+ raise RuntimeError(
+ f"Different parameters/ values for {tool}."
+ )
+
protocol.append(step)
# modelling step
# ToDo: Discuss input, seem to depend on mode
# ToDo: what about step details? Would it be nice to add the AlphaPulldown
# mode here?
+ m_type = "modeling"
step = {
- "method_type": "modeling",
+ "method_type": m_type,
"step_name": None,
"details": None,
"input_data_group": ["target_sequences", "STEPTYPE$coevolution MSA"],
"output_data_group": "model",
"software_group": ["AlphaPulldown", "AlphaFold"],
- "parameter_group": [
- sftwr["AlphaPulldown"]["params"],
- sftwr["AlphaFold"]["params"],
- ],
+ "parameter_group": [],
}
+ for sftwr in modelcif_json["__meta__"].values():
+ sftwr = sftwr["software"]
+ for i, tool in enumerate(["AlphaPulldown", "AlphaFold"]):
+ if i >= len(step["parameter_group"]):
+ step["parameter_group"].append(sftwr[tool][m_type])
+ else:
+ if step["parameter_group"][i] != sftwr[tool][m_type]:
+ raise RuntimeError(
+ f"Different parameters/ values for {tool}."
+ )
+
protocol.append(step)
# model selection step <- ask if there is automated selection, if only
--
GitLab