def get_spliceRack_bp(pseudocount=0.05):
return [
pseudocount +
PWM(np.loadtxt(BR_SPLICE_RACK_PATH + "/AT_AC_U12.txt"), "AT_AC_U12"),
pseudocount +
PWM(np.loadtxt(BR_SPLICE_RACK_PATH + "/GT_AG_U12.txt"), "GT_AG_U12")
]
def get_metadata():
"""
Get pandas.DataFrame with metadata about the PWM's. Columns:
- PWM_id (id of the PWM - pass to get_pwm_list() for getting the pwm
- TF
- Organism
- DB
- Info
- consensus
"""
motifs = load_motif_db(HOCOMOCO_PWM)
motif_names = sorted(list(motifs.keys()))
df = pd.Series(motif_names).str.split(pat="_|\\.", expand=True)
df.rename(columns={
0: "TF",
1: "Organism",
2: "DB",
3: "info"
},
inplace=True)
# add PWM_id
df.insert(0, "PWM_id", motif_names)
# compute the consensus
consensus = pd.Series(
[PWM(motifs[m]).get_consensus() for m in motif_names])
df["consensus"] = consensus
return df
def read_transfac(file_path, ignore_motif_name=True):
from concise.utils.pwm import PWM
from collections import defaultdict
with open(file_path) as f:
motif_lines = defaultdict(list)
if not ignore_motif_name:
motif = None
else:
motif = 0
for line in f:
if line.startswith("DE"):
if not ignore_motif_name:
motif = line.split(
"\t")[1].strip() # all are called motif 1
else:
motif += 1
elif line.startswith("XX"):
if not ignore_motif_name:
motif = None
else:
motif_lines[motif].append(line.split("\t"))
return {
str(motif): PWM(np.array(v)[:, 1:-1].astype(float) + 0.01)
for motif, v in motif_lines.items()
}
def __init__(self, pwm_list=[], stddev=0.05, seed=None):
if len(pwm_list) > 0 and isinstance(pwm_list[0], dict):
pwm_list = [PWM.from_config(pwm) for pwm in pwm_list]
self.stddev = stddev
self.seed = seed
self.pwm_list = pwm_list
_check_pwm_list(pwm_list)
def consensus_dist(seqs, seq_distance_fn, **kwargs):
from concise.utils.pwm import PWM
from concise.preprocessing.sequence import one_hot2string, DNA
consensus = PWM(seqs.mean(0) + 0.001).get_consensus()
return np.array([
seq_distance_fn(consensus, s, **kwargs)
for s in one_hot2string(seqs, DNA)
])
def read_meme_motifs(meme_file):
from basepair.external.meme import read
from concise.utils.pwm import PWM
with open(meme_file) as f:
record = read(f)
return {
str(i + 1): PWM(pd.DataFrame(m.pwm)[list("ACGT")].values + 0.01,
name=m.num_occurrences)
for i, m in enumerate(record)
}
def __init__(self, pwm_list=[], kernel_size=None, mean_max_scale=0.):
# handle pwm_list as a dictionary
if len(pwm_list) > 0 and isinstance(pwm_list[0], dict):
pwm_list = [PWM.from_config(pwm) for pwm in pwm_list]
if kernel_size is None:
kernel_size = len(pwm_list)
self.pwm_list = pwm_list
self.kernel_size = kernel_size
self.mean_max_scale = mean_max_scale
_check_pwm_list(pwm_list)
def get_branchpoint_pwm_list(cache=True):
l = []
if os.path.isfile(BR_PWM) and cache:
l.append(PWM.from_config(read_json(BR_PWM)))
else:
dt = pd.read_csv(
DATA_ROOT +
"/Splice_branchpoints/processed/branchpointer/train/filteredDescr.csv"
)
# colmeans
dt = dt[dt.set == "HC"]
dtseq = dt[dt.columns[dt.columns.str.match("^seq_")]] - 1
pwm = np.array(dtseq.mean()).reshape((-1, 4))
assert np.allclose(pwm.sum(1), 1)
p = PWM(pwm, name="U2_branchpoint")
write_json(p.get_config(), BR_PWM)
l.append(p)
l.append(
PWM(0.05 + np.loadtxt(BR_SPLICE_RACK_PATH + "/GT_AG_U12.txt"),
"GT_AG_U12_branchpoint"))
return l
def get_pwm_list(motif_name_list, pseudocountProb=0.0001):
"""Get a list of ENCODE PWM's.
# Arguments
pwm_id_list: List of id's from the `PWM_id` column in `get_metadata()` table
pseudocountProb: Added pseudocount probabilities to the PWM
# Returns
List of `concise.utils.pwm.PWM` instances.
"""
l = _load_motifs()
l = {k.split()[0]: v for k, v in l.items()}
pwm_list = [PWM(l[m] + pseudocountProb, name=m) for m in motif_name_list]
return pwm_list
def __init__(self,
pwm_list=[],
stddev=0.05,
seed=None,
background_probs=DEFAULT_BASE_BACKGROUND,
add_noise_before_Pwm2Pssm=True):
if len(pwm_list) > 0 and isinstance(pwm_list[0], dict):
pwm_list = [PWM.from_config(pwm) for pwm in pwm_list]
self.pwm_list = pwm_list
_check_pwm_list(pwm_list)
self.stddev = stddev
self.seed = seed
self.background_probs = background_probs
self.add_noise_before_Pwm2Pssm = add_noise_before_Pwm2Pssm
def get_metadata():
"""Get pandas.DataFrame with metadata about the PWM's. Columns:
- PWM_id (id of the PWM - pass to get_pwm_list() for getting the pwm
- info1 - additional information about the motifs
- info2
- consensus: PWM consensus sequence
"""
motifs = _load_motifs()
motif_names = sorted(list(motifs.keys()))
df = pd.Series(motif_names).str.split(expand=True)
df.rename(columns={0: "PWM_id", 1: "info1", 2: "info2"}, inplace=True)
# compute the consensus
consensus = pd.Series([PWM(motifs[m]).get_consensus() for m in motif_names])
df["consensus"] = consensus
return df
def __init__(self,
pwm_list=[],
kernel_size=None,
mean_max_scale=0.,
background_probs=DEFAULT_BASE_BACKGROUND):
# handle pwm_list as a dictionary
if len(pwm_list) > 0 and isinstance(pwm_list[0], dict):
pwm_list = [PWM.from_config(pwm) for pwm in pwm_list]
if kernel_size is None:
kernel_size = len(pwm_list)
_check_pwm_list(pwm_list)
self.pwm_list = pwm_list
self.kernel_size = kernel_size
self.mean_max_scale = mean_max_scale
self.background_probs = background_probs
def read_homer(file_path):
df = pd.read_csv(file_path, skiprows=1, header=None, sep='\t').values
l.split(",P:")[1].strip()
with open("motif1.motif.txt") as f:
l = f.readline()
return PWM(df + 0.01, name="P={}".format(l.split(",P:")[1].strip()))