def isotope_label_detector(
input_spec: InputSpec,
source_dir: Path,
exp_name: str,
min_scans: int = 5,
num_cond: int = 1,
n_jobs: int = -1,
):
scan_dir = source_dir.joinpath("all_scan_data")
slope_dir = source_dir.joinpath("all_slope_data")
slope_dir.mkdir(parents=True, exist_ok=True)
out_dir = source_dir.joinpath("all_isotope_analysis")
out_dir.mkdir(parents=True, exist_ok=True)
features = utils.get_featurelist(source_dir=source_dir, exp_name=exp_name)
conditions = input_spec.get_conditions()
def run_label_detector(cond):
out_file = slope_dir.joinpath(f"all_slope_data_{cond}.csv")
print(f"Detecting labels in {cond}")
#load both unlabeled and labeled scan data for cond
scan_files = list(scan_dir.glob(f"all_ions_*{cond}.csv"))
scan_dfs = []
for s in scan_files:
assert s.exists()
s_df = pd.read_csv(s)
scan_dfs.append(s_df)
df = utils.combine_dfs(scan_dfs)
#fil = scan_dir.joinpath(f"all_ions_{cond}.csv")
#assert fil.exists()
#df = pd.read_csv(fil)
grouped = df.groupby(["exp_id", "isotope", "condition"])
data = []
for (e_id, iso, c), g in grouped:
res = utils.calc_rep_stats(g, e_id, iso, c, min_scans=min_scans)
# print(data)
if len(res) < 1:
continue
data.extend(res)
if data:
agg_df = pd.DataFrame(data)
res_df = utils.aggregate_results(agg_df)
res_df.to_csv(out_file, index=False)
utils.run_label_analysis(res_df, cond, out_dir)
else:
print(f"No labels to detect for {cond}")
# Run processing of each condition in separate process
joblib.Parallel(n_jobs=n_jobs)(joblib.delayed(run_label_detector)(c)
for c in conditions)
# [run_label_detector(c) for c in conditions]
sum_df = utils.summarize_labels(out_dir, features, conditions)
sum_df.to_csv(source_dir.joinpath(f"{exp_name}_data_summary.csv"))
filtered_df = utils.filter_summary(sum_df, num_cond)
filtered_df.to_csv(
source_dir.joinpath(f"{exp_name}_data_summary_filtered.csv"))
def generate_featurelist(
input_spec: InputSpec,
source_dir: Path,
exp_name: str,
config: Dict,
n_jobs: int = -1,
blank_remove: bool = True,
):
"""
Create an mz ground truth list of all features detected in experiment and basketed
to align between conditions and replicates. Optional blank subtraction.
"""
print("Collecting feature list files")
# define here so input_spec and config in scope and not needed as params
def do_munge_featurelist(cond: str):
out_dir = source_dir.joinpath(cond)
out_dir.mkdir(parents=True, exist_ok=True)
print(f"Working on {cond}")
all_collapsed = utils.munge_featurelist(inp_spec=input_spec,
cond=cond,
out_dir=out_dir,
config=config)
return all_collapsed
# peaks in blanks to subtract later
# This should not break in the event that there are no blanks
# but the printing may be misleading
if input_spec.get_feature_filepaths("blank"):
blanks = do_munge_featurelist("blank")
else:
print("No blanks found")
blank_remove = False
# Run pre-processing on conditions in separate processes
conditions = input_spec.get_conditions()
cond_dfs = joblib.Parallel(n_jobs=n_jobs)(
joblib.delayed(do_munge_featurelist)(c) for c in conditions)
all_cond_df = utils.combine_dfs(cond_dfs)
if blank_remove:
print("Substracting blanks")
utils.blank_subtract(blanks, all_cond_df, config=config)
else:
print("Not subtracting blanks")
print("Grouping all features")
all_cond_df.reset_index(inplace=True, drop=True)
dereplicator.group_features(
all_cond_df, exp_name, "exp_id",
config=config) # final grouping - exp_id used for scan munging
all_cond_df.to_csv(source_dir.joinpath(f"{exp_name}_all_features.csv"),
index=False)
return all_cond_df
