def __init__(self, ncore=1, genome="hg38", gene_bed=None, pfmfile=None, include_notfs=False, rm_curated=True, etype="hg38H3K27ac", tffile=None):
self.ncore = ncore
self.genome = genome
# dream_model.txt is the logistic regression model.
package_dir = os.path.dirname(ananse.__file__)
self.etype = etype
if self.genome == "hg38" and self.etype == "hg38H3K27ac":
self.model = os.path.join(package_dir, "db", "dream_model_h3k27ac.txt")
elif self.etype == "p300" or self.etype == "ATAC":
self.model = os.path.join(package_dir, "db", "dream_model_p300.txt")
else:
raise TypeError("""The input enhancer data type should hg38H3K27ac, p300 or ATAC.
It is not possible set -e to hg38H3K27ac if the genome is not hg38.
Please provide a enhancer type with -e argument. By default is hg38H3K27ac.""")
# filter tfs?
self.include_notfs = include_notfs
# remove curated?
self.rm_curated = rm_curated
# load real tfs
self.tffile = tffile
if self.tffile is None:
self.tffile = os.path.join(package_dir, "db", "tfs.txt")
# self.tffile = "db/tfs.txt"
# Motif information file
self.pfmfile = pfmfile_location(pfmfile)
self.motifs2factors = self.pfmfile.replace(".pfm", ".motif2factors.txt")
self.filtermotifs2factors = clear_tfs(self.motifs2factors, self.tffile, self.include_notfs, self.rm_curated)
def __init__(
self,
peak_weights,
motif_weights,
pfmfile=None,
model=None,
curation_filter=None,
tf_list=None,
whitelist=True,
ncore=1,
verbose=True,
):
self.peak_weights = peak_weights # output from ScorePeaks
self.motif_weights = motif_weights # output from ScoreMotifs
self.motifs2factors_file = pfmfile_location(pfmfile).replace(
".pfm", ".motif2factors.txt")
self.motifs2factors = self.filter_transcription_factors(
curation_filter, tf_list, whitelist)
self.model = model
if self.model is None:
# dream_model.txt is a 2D logistic regression model.
package_dir = os.path.dirname(__file__)
self.model = os.path.join(package_dir, "db",
"dream_model_p300.pickle")
self.ncore = ncore
self.verbose = verbose
def logo(args):
if args.pfmfile is None and args.ids is None:
name = os.path.splitext(os.path.split(pfmfile_location(None))[-1])[0]
print(
"Use the -i argument to specify which motif ids you want to use for logos."
)
print("If you really want to create logos for all of the motifs in the default")
print("PFM file use the following command:")
print(f"gimme logo -p {name}")
sys.exit(1)
inputfile = args.pfmfile
motifs = read_motifs(inputfile)
if args.ids:
ids = args.ids.split(",")
motifs = [m for m in motifs if m.id in ids]
for motif in motifs:
motif.plot_logo(
fname="{}.png".format(motif.id), kind=args.kind, title=args.title
)
def moap(
inputfile,
method="hypergeom",
scoring=None,
outfile=None,
motiffile=None,
pfmfile=None,
genome=None,
fpr=0.01,
ncpus=None,
subsample=None,
zscore=True,
gc=True,
):
"""Run a single motif activity prediction algorithm.
Parameters
----------
inputfile : str
:1File with regions (chr:start-end) in first column and either cluster
name in second column or a table with values.
method : str, optional
Motif activity method to use. Any of 'hypergeom', 'lasso',
'bayesianridge',
'rf', 'xgboost'. Default is 'hypergeom'.
scoring: str, optional
Either 'score' or 'count'
outfile : str, optional
Name of outputfile to save the fitted activity values.
motiffile : str, optional
Table with motif scan results. First column should be exactly the same
regions as in the inputfile.
pfmfile : str, optional
File with motifs in pwm format. Required when motiffile is not
supplied.
genome : str, optional
Genome name, as indexed by gimme. Required when motiffile is not
supplied
fpr : float, optional
FPR for motif scanning
ncpus : int, optional
Number of threads to use. Default is the number specified in the config.
zscore : bool, optional
Use z-score normalized motif scores.
gc : bool optional
Use GC% bins for z-score.
Returns
-------
pandas DataFrame with motif activity
"""
if scoring and scoring not in ["score", "count"]:
raise ValueError("valid values are 'score' and 'count'")
if inputfile.endswith("feather"):
df = pd.read_feather(inputfile)
df = df.set_index(df.columns[0])
else:
# read data
df = pd.read_table(inputfile, index_col=0, comment="#")
clf = Moap.create(method, ncpus=ncpus)
if clf.ptype == "classification":
if df.shape[1] != 1:
raise ValueError("1 column expected for {}".format(method))
else:
if np.dtype("object") in set(df.dtypes):
raise ValueError(
"columns should all be numeric for {}".format(method))
if motiffile is None:
if genome is None:
raise ValueError("need a genome")
pfmfile = pfmfile_location(pfmfile)
try:
motifs = read_motifs(pfmfile)
except Exception:
sys.stderr.write("can't read motifs from {}".format(pfmfile))
raise
# scan for motifs
motif_names = [m.id for m in read_motifs(pfmfile)]
scores = []
if method == "classic" or scoring == "count":
logger.info("motif scanning (scores)")
scores = scan_regionfile_to_table(
inputfile,
genome,
"count",
pfmfile=pfmfile,
ncpus=ncpus,
zscore=zscore,
gc=gc,
)
else:
logger.info("motif scanning (scores)")
scores = scan_regionfile_to_table(
inputfile,
genome,
"score",
pfmfile=pfmfile,
ncpus=ncpus,
zscore=zscore,
gc=gc,
)
motifs = pd.DataFrame(scores, index=df.index, columns=motif_names)
elif isinstance(motiffile, pd.DataFrame):
motifs = motiffile
else:
motifs = pd.read_table(motiffile, index_col=0, comment="#")
if outfile and os.path.exists(outfile):
out = pd.read_table(outfile, index_col=0, comment="#")
ncols = df.shape[1]
if ncols == 1:
ncols = len(df.iloc[:, 0].unique())
if out.shape[0] == motifs.shape[1] and out.shape[1] == ncols:
logger.warn("%s output already exists... skipping", method)
return out
if subsample is not None:
n = int(subsample * df.shape[0])
logger.debug("Subsampling %d regions", n)
df = df.sample(n)
motifs = motifs.loc[df.index]
clf.fit(motifs, df)
if outfile:
with open(outfile, "w") as f:
f.write(
"# maelstrom - GimmeMotifs version {}\n".format(__version__))
f.write("# method: {} with motif {}\n".format(method, scoring))
if genome:
f.write("# genome: {}\n".format(genome))
if isinstance(motiffile, str):
f.write("# motif table: {}\n".format(motiffile))
f.write("# {}\n".format(clf.act_description))
with open(outfile, "a") as f:
clf.act_.to_csv(f, sep="\t")
return clf.act_
def infer_motifs(
adata: AnnData,
dataset: str,
cluster: Optional[str] = "louvain",
n_top_genes: Optional[int] = 1000,
max_cell_types: Optional[int] = 50,
pfm: Optional[str] = None,
min_annotated: Optional[int] = 50,
num_enhancers: Optional[int] = 10000,
maelstrom: Optional[bool] = False,
indirect: Optional[bool] = True,
n_sketch: Optional[int] = 2500,
n_permutations: Optional[int] = 100000,
) -> None:
"""Infer motif ativity for single cell RNA-seq data.
The adata object is modified with the following fields.
**X_cell_types** : `adata.obsm` field
Cell type coefficients.
Parameters
----------
adata : :class:`~anndata.AnnData`
Annotated data matrix.
dataset : `str`
Name of reference data set or directory with reference data.
cluster : `str`, optional (default: "louvain")
Name of the clustering, can be either louvain or leiden.
n_top_genes : `int`, optional (default: 1000)
Number of variable genes that is used. If `n_top_genes` is greater than the
number of hypervariable genes in `adata` then all variable genes are
used.
max_cell_types : `int`, optional (default: 50)
Maximum number of cell types to select.
pfm : `str`, optional (default: None)
Name of motif file in PFM format. The GimmeMotifs default is used
if this parameter is not specified. This can be a filename, or a
pfm name support by GimmeMotifs such as `JASPAR2018_vertebrates`.
If a custom PFM file is specified, there should also be an associated
`.motif2factors.txt` file.
min_annotated : `int`, optional (default: 50)
Cells that are annotated with cell types less than this number will be
annotated as "other".
num_enhancers : `int`, optional (default: 10000)
Number of enhancers to use for motif activity analysis.
maelstrom : `boolean`, optional (default: False)
Use maelstrom instead of ridge regression for motif activity analysis.
"""
use_name = True
validate_adata(adata)
data = ScepiaDataset(dataset)
if "scepia" not in adata.uns:
adata.uns["scepia"] = {"version": __version__}
# Annotate each cell with H3K27ac reference
if "cell_annotation" not in adata.obs or "cluster_annotation" not in adata.obs:
annotate_cells(
adata,
dataset=dataset,
cluster=cluster,
n_top_genes=n_top_genes,
min_annotated=min_annotated,
max_cell_types=max_cell_types,
)
logger.info("Linking variable genes to differential enhancers.")
gene_map_file = data.gene_mapping
link_file = data.link_file
link = pd.read_feather(link_file)
if use_name:
ens2name = pd.read_csv(gene_map_file,
sep="\t",
index_col=0,
names=["identifier", "name"])
link = link.join(ens2name, on="gene").dropna()
link = link.set_index("name")
link.index = link.index.str.upper()
enh_genes = adata.var_names[adata.var_names.str.upper().isin(
link.index)].str.upper()
var_enhancers = change_region_size(link.loc[enh_genes, "loc"]).unique()
enhancer_df = data.load_reference_data(reftype="enhancer")
enhancer_df.index = change_region_size(enhancer_df.index)
enhancer_df = enhancer_df.loc[var_enhancers,
adata.uns["scepia"]["cell_types"]]
enhancer_df = enhancer_df.groupby(enhancer_df.columns, axis=1).mean()
enhancer_df.loc[:, :] = scale(enhancer_df)
# Select top most variable enhancers
enhancer_df = enhancer_df.loc[enhancer_df.var(1).sort_values().tail(
num_enhancers).index]
# Center by mean of the most import cell types
# Here we chose the majority cell type per cluster
cluster_cell_types = adata.obs["cluster_annotation"].unique()
mean_value = enhancer_df[cluster_cell_types].mean(1)
enhancer_df = enhancer_df.sub(mean_value, axis=0)
fname = NamedTemporaryFile(delete=False).name
enhancer_df.to_csv(fname, sep="\t")
logger.info("inferring motif activity")
pfm = pfmfile_location(pfm)
if maelstrom:
with TemporaryDirectory() as tmpdir:
run_maelstrom(
fname,
data.genome,
tmpdir,
center=False,
filter_redundant=True,
)
motif_act = pd.read_csv(
os.path.join(tmpdir, "final.out.txt"),
sep="\t",
comment="#",
index_col=0,
)
motif_act.columns = motif_act.columns.str.replace(
r"z-score\s+", "")
pfm = pfmfile_location(
os.path.join(tmpdir, "nonredundant.motifs.pfm"))
else:
logger.info(f"Activity based on genome {data.genome}")
motif_act = moap(
fname,
scoring="score",
genome=data.genome,
method="bayesianridge",
pfmfile=pfm,
ncpus=12,
)
adata.uns["scepia"]["pfm"] = pfm
adata.uns["scepia"]["motif_activity"] = motif_act[adata.uns["scepia"]
["cell_types"]]
logger.info("calculating cell-specific motif activity")
cell_motif_activity = (
adata.uns["scepia"]["motif_activity"] @ adata.obsm["X_cell_types"].T).T
cell_motif_activity.index = adata.obs_names
adata.obs = adata.obs.drop(
columns=cell_motif_activity.columns.intersection(adata.obs.columns))
adata.obs = adata.obs.join(cell_motif_activity)
correlate_tf_motifs(adata,
indirect=indirect,
n_sketch=n_sketch,
n_permutations=n_permutations)
add_activity(adata)
logger.info("Done with motif inference.")
def select_nonredundant_motifs(roc_report,
pfmfile,
fg_table,
bg_table,
tolerance=0.001):
pfmfile = pfmfile_location(pfmfile)
motifs = read_motifs(pfmfile)
motif_dict = read_motifs(pfmfile, as_dict=True)
mc = MotifComparer()
df = pd.read_csv(roc_report, sep="\t", index_col=0)
df = df[df["Enr. at 1% FPR"] >= 2]
motifs = [m for m in motifs if m.id in df.index]
cols = ["ROC AUC", "PR AUC", "Enr. at 1% FPR", "Recall at 10% FDR"]
rank = df[cols].rank().mean(1).sort_values(ascending=False)
redundant_motifs = []
keep = []
while df[~df.index.isin(redundant_motifs)].shape[0] > 0:
motif = rank[~rank.index.isin(redundant_motifs)].head(1).index[0]
keep.append(motif)
result = mc.get_all_scores(
[motif_dict[motif]],
[m for m in motifs if m.id not in redundant_motifs],
"partial",
"seqcor",
"mean",
)
result = result[motif]
redundant_motifs += [m for m in result.keys() if result[m][0] >= 0.7]
logger.debug(f"Selected {len(keep)} motifs for feature elimination")
# Read motif scan results
fg_table = pd.read_csv(fg_table, index_col=0, comment="#", sep="\t")
bg_table = pd.read_csv(bg_table, index_col=0, comment="#", sep="\t")
X = pd.concat((fg_table, bg_table), axis=0)
y = np.hstack((np.ones(fg_table.shape[0]), np.zeros(bg_table.shape[0])))
X_train, X_test, y_train, y_test = train_test_split(
X,
y,
test_size=0.4,
random_state=2,
shuffle=True,
)
X_bla = X_train[keep]
model = LogisticRegression(solver="liblinear", max_iter=500, penalty="l1")
# = RandomForestClassifier(n_estimators=100)
max_score = np.mean(
cross_val_score(model,
X_bla,
y_train,
cv=5,
scoring="average_precision"))
mean_scores = []
step = 1
logger.info("selecting non-redundant motifs")
n_features = 1
for i in range(1, X_bla.shape[1], step):
rfe = RFE(model, i)
fit = rfe.fit(X_bla, y_train)
mean_score = np.mean(
cross_val_score(
model,
X_bla.loc[:, fit.support_],
y_train,
cv=5,
scoring="average_precision",
))
if i > 1 and mean_score - mean_scores[-1] < (max_score * tolerance):
n_features = i - 1
break
mean_scores.append(mean_score)
rfe = RFE(model, n_features)
fit = rfe.fit(X_bla, y_train)
selected_features = X_bla.columns[fit.support_]
model.fit(X_train.loc[:, selected_features], y_train)
y_pred = model.predict_proba(X_test.loc[:, selected_features])[:, 1]
pr_auc = average_precision_score(y_test, y_pred)
roc_auc = roc_auc_score(y_test, y_pred)
logger.info(
f"selected {len(selected_features)} non-redundant motifs: ROC AUC {roc_auc:.3f}, PR AUC {pr_auc:.3f}"
)
return selected_features
def run_maelstrom(
infile,
genome,
outdir,
pfmfile=None,
plot=True,
cluster=False,
score_table=None,
count_table=None,
methods=None,
ncpus=None,
zscore=True,
gc=True,
):
"""Run maelstrom on an input table.
Parameters
----------
infile : str
Filename of input table. Can be either a text-separated tab file or a
feather file.
genome : str
Genome name. Can be either the name of a FASTA-formatted file or a
genomepy genome name.
outdir : str
Output directory for all results.
pfmfile : str, optional
Specify a PFM file for scanning.
plot : bool, optional
Create heatmaps.
cluster : bool, optional
If True and if the input table has more than one column, the data is
clustered and the cluster activity methods are also run. Not
well-tested.
score_table : str, optional
Filename of pre-calculated table with motif scores.
count_table : str, optional
Filename of pre-calculated table with motif counts.
methods : list, optional
Activity methods to use. By default are all used.
ncpus : int, optional
If defined this specifies the number of cores to use.
zscore : bool, optional
Use z-score normalized motif scores.
gc : bool, optional
Use GC% bins to normalize motif scores.
"""
logger.info("Starting maelstrom")
if infile.endswith("feather"):
df = pd.read_feather(infile)
df = df.set_index(df.columns[0])
else:
df = pd.read_table(infile, index_col=0, comment="#")
# Check for duplicates
if df.index.duplicated(keep=False).any():
logger.warning("Input file contains duplicate regions!")
logger.warning("These will be removed.")
df = df.iloc[~df.index.duplicated(keep=False)]
if not os.path.exists(outdir):
os.mkdir(outdir)
if methods is None:
methods = Moap.list_predictors()
methods = [m.lower() for m in methods]
df.to_csv(os.path.join(outdir, "input.table.txt"), sep="\t")
infile = os.path.join(outdir, "input.table.txt")
# Copy the motif informatuon
pfmfile = pfmfile_location(pfmfile)
if pfmfile:
shutil.copy2(pfmfile, outdir)
mapfile = re.sub(".p[fw]m$", ".motif2factors.txt", pfmfile)
if os.path.exists(mapfile):
shutil.copy2(mapfile, outdir)
# Create a file with the number of motif matches
if count_table is None:
count_table = os.path.join(outdir, "motif.count.txt.gz")
if not os.path.exists(count_table):
logger.info("motif scanning (counts)")
counts = scan_to_table(
infile,
genome,
"count",
pfmfile=pfmfile,
ncpus=ncpus,
zscore=zscore,
gc=gc,
)
counts.to_csv(count_table, sep="\t", compression="gzip")
else:
logger.info("Counts, using: %s", count_table)
# Create a file with the score of the best motif match
if score_table is None:
score_table = os.path.join(outdir, "motif.score.txt.gz")
if not os.path.exists(score_table):
logger.info("motif scanning (scores)")
scores = scan_to_table(
infile,
genome,
"score",
pfmfile=pfmfile,
ncpus=ncpus,
zscore=zscore,
gc=gc,
)
scores.to_csv(score_table,
sep="\t",
float_format="%.3f",
compression="gzip")
else:
logger.info("Scores, using: %s", score_table)
if cluster:
cluster = False
for method in methods:
m = Moap.create(method, ncpus=ncpus)
if m.ptype == "classification":
cluster = True
break
if not cluster:
logger.info("Skipping clustering, no classification methods")
exps = []
clusterfile = infile
if df.shape[1] != 1:
# More than one column
for method in Moap.list_regression_predictors():
if method in methods:
m = Moap.create(method, ncpus=ncpus)
exps.append([method, m.pref_table, infile])
logger.debug("Adding %s", method)
if cluster:
clusterfile = os.path.join(
outdir,
os.path.basename(infile) + ".cluster.txt")
df[:] = scale(df, axis=0)
names = df.columns
df_changed = pd.DataFrame(index=df.index)
df_changed["cluster"] = np.nan
for name in names:
df_changed.loc[(df[name] -
df.loc[:, df.columns != name].max(1)) > 0.5,
"cluster"] = name
df_changed.dropna().to_csv(clusterfile, sep="\t")
if df.shape[1] == 1 or cluster:
for method in Moap.list_classification_predictors():
if method in methods:
m = Moap.create(method, ncpus=ncpus)
exps.append([method, m.pref_table, clusterfile])
if len(exps) == 0:
logger.error("No method to run.")
sys.exit(1)
for method, scoring, fname in exps:
try:
if scoring == "count" and count_table is not None:
moap_with_table(fname,
count_table,
outdir,
method,
scoring,
ncpus=ncpus)
elif scoring == "score" and score_table is not None:
moap_with_table(fname,
score_table,
outdir,
method,
scoring,
ncpus=ncpus)
else:
moap_with_bg(fname,
genome,
outdir,
method,
scoring,
pfmfile=pfmfile,
ncpus=ncpus)
except Exception as e:
logger.warn("Method %s with scoring %s failed", method, scoring)
logger.warn(e)
logger.warn("Skipping")
raise
dfs = {}
for method, scoring, fname in exps:
t = "{}.{}".format(method, scoring)
fname = os.path.join(outdir,
"activity.{}.{}.out.txt".format(method, scoring))
try:
dfs[t] = pd.read_table(fname, index_col=0, comment="#")
except FileNotFoundError:
logger.warn("Activity file for {} not found!\n".format(t))
if len(methods) > 1:
logger.info("Rank aggregation")
df_p = df_rank_aggregation(df, dfs, exps)
df_p.to_csv(os.path.join(outdir, "final.out.txt"), sep="\t")
# df_p = df_p.join(m2f)
# Write motif frequency table
if df.shape[1] == 1:
mcount = df.join(pd.read_table(count_table, index_col=0, comment="#"))
m_group = mcount.groupby(df.columns[0])
freq = m_group.sum() / m_group.count()
freq.to_csv(os.path.join(outdir, "motif.freq.txt"), sep="\t")
if plot and len(methods) > 1:
logger.info("html report")
maelstrom_html_report(outdir, os.path.join(outdir, "final.out.txt"),
pfmfile)
logger.info(os.path.join(outdir, "gimme.maelstrom.report.html"))
def calc_stats_iterator(
fg_file=None,
bg_file=None,
fg_table=None,
bg_table=None,
motifs=None,
stats=None,
genome=None,
zscore=True,
gc=True,
ncpus=None,
):
"""Calculate motif enrichment metrics.
Parameters
----------
fg_file : str, optional
Filename of a FASTA, BED or region file with positive sequences.
bg_file : str, optional
Filename of a FASTA, BED or region file with negative sequences.
fg_table : str, optional
Filename of a table with motif scan results of positive sequences.
bg_table : str, optional
Filename of a table with motif scan results of negative sequences.
motifs : str, list or Motif instance, optional
A file with motifs in pfm format, a list of Motif instances or a
single Motif instance. If motifs is `None`, the default motif
database is used.
genome : str, optional
Genome or index directory in case of BED/regions.
stats : list, optional
Names of metrics to calculate. See gimmemotifs.rocmetrics.__all__
for available metrics.
ncpus : int, optional
Number of cores to use.
Returns
-------
result : dict
Dictionary with results where keys are motif ids and the values are
dictionary with metric name and value pairs.
"""
if not stats:
stats = rocmetrics.__all__
if fg_table is None:
if fg_file is None:
raise ValueError("Need either fg_table or fg_file argument")
elif fg_file is not None:
raise ValueError("Need either fg_table or fg_file argument, not both")
if bg_table is None:
if bg_file is None:
raise ValueError("Need either bg_table or bg_file argument")
elif bg_file is not None:
raise ValueError("Need either bg_table or bg_file argument, not both")
if fg_table is not None or bg_table is not None:
remove_stats = []
for s in stats:
func = getattr(rocmetrics, s)
if func.input_type == "pos":
remove_stats.append(s)
if len(remove_stats) != 0:
logger.warn(
"Cannot calculate stats that require position from table of motif scores."
)
logger.warn(f"Skipping the following statistics: {', '.join(remove_stats)}")
stats = [s for s in stats if s not in remove_stats]
if isinstance(motifs, Motif):
all_motifs = [motifs]
else:
if type([]) == type(motifs):
all_motifs = motifs
else:
motifs = pfmfile_location(motifs)
all_motifs = read_motifs(motifs, fmt="pwm")
if fg_table is not None or bg_table is not None:
filtered_motifs = pd.read_csv(
fg_table, sep="\t", index_col=0, nrows=1, comment="#"
).columns
filtered_motifs = filtered_motifs.intersection(
pd.read_csv(bg_table, sep="\t", index_col=0, nrows=1, comment="#").columns
)
all_motifs = [m for m in all_motifs if m.id in filtered_motifs]
if ncpus is None:
ncpus = int(MotifConfig().get_default_params()["ncpus"])
if fg_file is not None or bg_file is not None:
if zscore or gc:
# Precalculate mean and stddev for z-score calculation
s = Scanner(ncpus=ncpus)
s.set_motifs(all_motifs)
s.set_genome(genome)
s.set_meanstd(gc=gc)
chunksize = 240
for i in range(0, len(all_motifs), chunksize):
result = {}
logger.debug(
"chunk %s of %s", (i / chunksize) + 1, len(all_motifs) // chunksize + 1
)
motifs = all_motifs[i : i + chunksize]
if fg_table is None:
fg_total = scan_to_best_match(
fg_file, motifs, ncpus=ncpus, genome=genome, zscore=zscore, gc=gc
)
else:
fg_total = pd.read_csv(
fg_table, sep="\t", usecols=[m.id for m in motifs], comment="#"
).to_dict(orient="list")
for m in fg_total:
fg_total[m] = [(x, None) for x in fg_total[m]]
if bg_table is None:
bg_total = scan_to_best_match(
bg_file, motifs, ncpus=ncpus, genome=genome, zscore=zscore, gc=gc
)
else:
bg_total = pd.read_csv(
bg_table, sep="\t", usecols=[m.id for m in motifs], comment="#"
).to_dict(orient="list")
for m in bg_total:
bg_total[m] = [(x, None) for x in bg_total[m]]
logger.debug("calculating statistics")
if ncpus == 1:
it = _single_stats(motifs, stats, fg_total, bg_total)
else:
it = _mp_stats(motifs, stats, fg_total, bg_total, ncpus)
for motif_id, s, ret in it:
if motif_id not in result:
result[motif_id] = {}
result[motif_id][s] = ret
yield result
def __init__(self, genome, bed, pfmfile=None, ncore=1, verbose=True):
self.genome = genome
self.bed = bed # putative enhancer regions in format chr:start-end (in column 0 with header)
self.pfm_file = pfmfile_location(pfmfile)
self.ncore = ncore
self.verbose = verbose
def run_maelstrom(
infile,
genome,
outdir,
pfmfile=None,
filter_redundant=True,
filter_cutoff=0.8,
plot=True,
cluster=False,
score_table=None,
count_table=None,
methods=None,
ncpus=None,
zscore=True,
gc=True,
center=False,
aggregation="int_stouffer",
):
"""Run maelstrom on an input table.
Parameters
----------
infile : str
Filename of input table. Can be either a text-separated tab file or a
feather file.
genome : str
Genome name. Can be either the name of a FASTA-formatted file or a
genomepy genome name.
outdir : str
Output directory for all results.
pfmfile : str, optional
Specify a PFM file for scanning.
filter_redundant : bool, optional
Create a non-redundant set of motifs based on correlation of motif scores in the input data.
filter_cutoff : float, optional
Cutoff to use for non-redundant motif selection. Default is 0.8.
plot : bool, optional
Create heatmaps.
cluster : bool, optional
If True and if the input table has more than one column, the data is
clustered and the cluster activity methods are also run. Not
well-tested.
score_table : str, optional
Filename of pre-calculated table with motif scores.
count_table : str, optional
Filename of pre-calculated table with motif counts.
methods : list, optional
Activity methods to use. By default are all used.
ncpus : int, optional
If defined this specifies the number of cores to use.
zscore : bool, optional
Use z-score normalized motif scores.
gc : bool, optional
Use GC% bins to normalize motif scores.
center : bool, optional
Mean-center the input table.
aggregation: str, optional
How to combine scores of the predictors. The default is "int_stouffer", for
inverse normal transform followed by Stouffer's methods to combine z-scores.
Alternatively, "stuart" performs rank aggregation and reports the -log10 of
the rank aggregation p-value.
"""
logger.info("Starting maelstrom")
if infile.endswith("feather"):
df = pd.read_feather(infile)
df = df.set_index(df.columns[0])
else:
df = pd.read_table(infile, index_col=0, comment="#")
# Check if the input is mean-centered
if df.shape[1] > 1 and not np.allclose(df.mean(1), 0):
if center:
logger.info(
"Input is not mean-centered, setting the mean of all rows to 0."
)
logger.info(
"Use --nocenter if you know what you're doing and want to change this behavior."
)
logger.info(
"Note that if you use count data (ChIP-seq, ATAC-seq) we recommend to "
"first transform your data, for instance using log2(), and to normalize "
"between samples. To create a table suitable for maelstrom you can use the "
"coverage_table script included with GimmeMotifs."
)
df = df.sub(df.mean(axis=1), axis=0)
else:
logger.info("Input is not mean-centered, but --nocenter was specified.")
logger.info(
"Leaving the data as-is, but make sure this is what your really want."
)
# Check for duplicates
if df.index.duplicated(keep=False).any():
logger.warning("Input file contains duplicate regions!")
logger.warning("These will be removed.")
df = df.iloc[~df.index.duplicated(keep=False)]
if not os.path.exists(outdir):
os.mkdir(outdir)
if methods is None:
methods = Moap.list_predictors()
methods = [m.lower() for m in methods]
df.to_csv(os.path.join(outdir, "input.table.txt"), sep="\t")
infile = os.path.join(outdir, "input.table.txt")
# Copy the motif informatuon
pfmfile = pfmfile_location(pfmfile)
if pfmfile:
shutil.copy2(pfmfile, outdir)
mapfile = re.sub(".p[fw]m$", ".motif2factors.txt", pfmfile)
if os.path.exists(mapfile):
shutil.copy2(mapfile, outdir)
# Create a file with the number of motif matches
if count_table is None:
count_table = os.path.join(outdir, "motif.count.txt.gz")
if not os.path.exists(count_table):
logger.info("motif scanning (counts)")
counts = scan_regionfile_to_table(
infile,
genome,
"count",
pfmfile=pfmfile,
ncpus=ncpus,
zscore=zscore,
gc=gc,
)
counts.to_csv(count_table, sep="\t", compression="gzip")
else:
logger.info("Counts, using: %s", count_table)
# Create a file with the score of the best motif match
if score_table is None:
score_table = os.path.join(outdir, "motif.score.txt.gz")
if not os.path.exists(score_table):
logger.info("motif scanning (scores)")
scores = scan_regionfile_to_table(
infile,
genome,
"score",
pfmfile=pfmfile,
ncpus=ncpus,
zscore=zscore,
gc=gc,
)
scores.to_csv(
score_table, sep="\t", float_format="%.3f", compression="gzip"
)
else:
logger.info("Scores, using: %s", score_table)
counts = pd.read_csv(count_table, index_col=0, comment="#", sep="\t")
scores = pd.read_csv(score_table, index_col=0, comment="#", sep="\t")
if filter_redundant:
logger.info("Selecting non-redundant motifs")
fa = FeatureAgglomeration(
distance_threshold=filter_cutoff,
n_clusters=None,
affinity="correlation",
linkage="complete",
compute_full_tree=True,
)
fa.fit(scores)
X_cluster = pd.DataFrame({"motif": scores.columns, "label": fa.labels_})
X_cluster = X_cluster.join(scores.var().to_frame(name="var"), on="motif")
selected_motifs = (
X_cluster.sort_values("var")
.drop_duplicates(subset=["label"], keep="last")["motif"]
.values
)
nr_motif = (
X_cluster.sort_values("var")
.drop_duplicates(subset=["label"], keep="last")[["label", "motif"]]
.set_index("label")
)
X_cluster = X_cluster.join(nr_motif, rsuffix="_nr", on="label")
motif_map = X_cluster[["motif", "motif_nr"]].set_index("motif")
scores = scores[selected_motifs]
counts = counts[selected_motifs]
score_table = os.path.join(outdir, "motif.nr.score.txt.gz")
scores.to_csv(score_table, sep="\t", compression="gzip")
count_table = os.path.join(outdir, "motif.nr.count.txt.gz")
counts.to_csv(count_table, sep="\t", compression="gzip")
m2f = pd.read_table(os.path.join(outdir, mapfile), comment="#")
m2f = m2f.join(motif_map, on="Motif")
m2f.loc[m2f["Motif"] != m2f["motif_nr"], "Curated"] = "N"
m2f["Motif"] = m2f["motif_nr"]
m2f = m2f.drop(columns=["motif_nr"])
motifs = read_motifs(pfmfile)
pfmfile = os.path.join(outdir, "nonredundant.motifs.pfm")
with open(pfmfile, "w") as f:
for motif in motifs:
f.write(f"{motif.to_pfm()}\n")
mapfile = pfmfile.replace(".pfm", ".motif2factors.txt")
with open(mapfile, "w") as f:
f.write(
"# Note: this mapping is specifically created for this non-redundant set of motifs.\n"
)
f.write(
"# It also includes factors for motifs that were similar, but this can be\n"
)
f.write("# specific to this analysis.\n")
with open(mapfile, "a") as f:
m2f.to_csv(f, index=False, sep="\t")
logger.info(f"Selected {len(selected_motifs)} motifs")
logger.info(f"Motifs: {pfmfile}")
logger.info(f"Factor mappings: {mapfile}")
if cluster:
cluster = False
for method in methods:
m = Moap.create(method, ncpus=ncpus)
if m.ptype == "classification":
cluster = True
break
if not cluster:
logger.info("Skipping clustering, no classification methods")
exps = []
clusterfile = infile
if df.shape[1] != 1:
# More than one column
for method in Moap.list_regression_predictors():
if method in methods:
m = Moap.create(method, ncpus=ncpus)
exps.append([method, m.pref_table, infile])
logger.debug("Adding %s", method)
if cluster:
clusterfile = os.path.join(
outdir, os.path.basename(infile) + ".cluster.txt"
)
df[:] = scale(df, axis=0)
names = df.columns
df_changed = pd.DataFrame(index=df.index)
df_changed["cluster"] = np.nan
for name in names:
df_changed.loc[
(df[name] - df.loc[:, df.columns != name].max(1)) > 0.5, "cluster"
] = name
df_changed.dropna().to_csv(clusterfile, sep="\t")
if df.shape[1] == 1 or cluster:
for method in Moap.list_classification_predictors():
if method in methods:
m = Moap.create(method, ncpus=ncpus)
exps.append([method, m.pref_table, clusterfile])
if len(exps) == 0:
logger.error("No method to run.")
sys.exit(1)
for method, scoring, fname in exps:
try:
if scoring == "count":
moap_with_table(
fname, count_table, outdir, method, scoring, ncpus=ncpus
)
elif scoring == "score":
moap_with_table(
fname, score_table, outdir, method, scoring, ncpus=ncpus
)
except Exception as e:
logger.warn("Method %s with scoring %s failed", method, scoring)
logger.warn(e)
logger.warn("Skipping")
raise
dfs = {}
for method, scoring, fname in exps:
t = "{}.{}".format(method, scoring)
fname = os.path.join(outdir, "activity.{}.{}.out.txt".format(method, scoring))
try:
dfs[t] = pd.read_table(fname, index_col=0, comment="#")
except FileNotFoundError:
logger.warn("Activity file for {} not found!\n".format(t))
if len(methods) > 1:
logger.info("Rank aggregation")
df_p = df_rank_aggregation(df, dfs, exps, method=aggregation)
# Add percentage of input sequences with motif
if df.shape[1] > 1:
df_p["% with motif"] = counts[df_p.index].sum(0) / df.shape[0] * 100
else:
bla = counts.join(df).groupby(df.columns[0]).mean() * 100
bla = bla.T
bla = bla.rename(
columns={col: f"{col} % with motif" for col in bla.columns}
)
df_p = df_p.join(bla)
if df.shape[1] > 1:
# Add correlation between motif score and signal
logger.info("Correlation")
for col in df.columns:
df_p[f"corr {col}"] = 0
for motif in df_p.index:
df_p.loc[motif, f"corr {col}"] = pearsonr(df[col], scores[motif])[0]
df_p.to_csv(os.path.join(outdir, "final.out.txt"), sep="\t")
# df_p = df_p.join(m2f)
# Write motif frequency table
if df.shape[1] == 1:
mcount = df.join(counts)
m_group = mcount.groupby(df.columns[0])
freq = m_group.sum() / m_group.count()
freq.to_csv(os.path.join(outdir, "motif.freq.txt"), sep="\t")
if plot and len(methods) > 1:
logger.info("html report")
maelstrom_html_report(outdir, os.path.join(outdir, "final.out.txt"), pfmfile)
logger.info(os.path.join(outdir, "gimme.maelstrom.report.html"))
