Export files¶
In [ ]:
# Imports
from basepair.imports import *
In [22]:
dfi_subset_pr = pr.PyRanges(dfi_subset)
In [23]:
# duplicated number of instances
len(dfi_subset_pr)
Out[23]:
In [24]:
# non-duplicated number of instances
len(dfi_subset_pr.drop_duplicate_positions())
Out[24]:
In [25]:
dfi_subset.head()
Out[25]:
In [26]:
dict(dfi_subset.pattern_name.value_counts())
Out[26]:
In [27]:
fig, ax = plt.subplots(figsize=get_figsize(0.23))
dfi_subset.pattern_name.value_counts().plot.bar(width=.9)
plt.ylabel("Motif instances")
plt.xlabel("Motif")
fig.savefig(fdir / '../n-instances-per-motif.pdf')
In [103]:
# Number of regions with a motif (total = 150k. Only 14k lack a motif)
len(dfi_subset.groupby('example_idx').size())
Out[103]:
In [104]:
dfi_subset.groupby('example_idx').size().value_counts()
Out[104]:
In [113]:
total_regions = len(ranges)
In [114]:
fig, ax = plt.subplots(figsize=get_figsize(0.4))
vec = dfi_subset.groupby('example_idx').size()
vec = pd.concat([vec, pd.Series(np.zeros(total_regions - len(vec)))])
vec.plot.hist(bins=np.arange(11), rwidth=0.9, align='left')
plt.xlabel("Number of motifs per region")
fig.savefig(fdir / '../n-motifs-per-region.pdf')
In [16]:
dfi_subset.head()
Out[16]:
In [198]:
19.535 / (212.63-196.32) * 0.3
Out[198]:
In [199]:
3.505 / (212.63-196.32) * 0.3
Out[199]:
In [18]:
dfi_subset.groupby(['pattern_name', 'example_idx']).size().reset_index()
Out[18]:
In [ ]:
dfi_subset.head()
In [120]:
fig, axes = plt.subplots(len(tasks), 1,
figsize=get_figsize(0.23*1.4, aspect=len(tasks)* 0.618/1.4),
sharey=True,
gridspec_kw=dict(hspace=0),
sharex=True)
for i, task in enumerate(tasks):
ax = axes[i]
vec = dfi_subset[dfi_subset.example_interval_from_task == task].groupby('example_idx').size()
vec = pd.concat([vec, pd.Series(np.zeros(peaks_per_task[task] - vec.index.nunique()))])
vec.plot.hist(bins=np.arange(10), rwidth=0.9, align='left', ax=ax)
ax.set_ylabel(task, rotation=0, labelpad=10, ha='right')
ax.set_xlabel("Number of motifs per region")
fig.savefig(fdir / '../n-motifs-per-region.per-task.pdf')
In [121]:
fig, axes = plt.subplots(len(tasks), 1,
figsize=get_figsize(0.23, aspect=len(tasks)* 0.618),
sharey=True,
gridspec_kw=dict(hspace=0),
sharex=True)
pattern_names = dfi_subset.pattern_name.unique()
for i, task in enumerate(tasks):
ax = axes[i]
if i == 0:
ax.set_title("Total number of motifs")
vec = dfi_subset[dfi_subset.example_interval_from_task == task].pattern_name.value_counts().reindex(pattern_names, fill_value=0)
vec.plot.bar(width=.9, ax=ax)
ax.set_ylabel(task, rotation=0, labelpad=10, ha='right')
ax.set_xlabel("Motif")
fig.savefig(fdir / '../n-instances-per-motif.per-task.pdf')
In [122]:
fig, axes = plt.subplots(len(tasks), 1,
figsize=get_figsize(0.23, aspect=len(tasks)* 0.618),
sharey=True,
gridspec_kw=dict(hspace=0),
sharex=True)
pattern_names = dfi_subset.pattern_name.unique()
for i, task in enumerate(tasks):
ax = axes[i]
if i == 0:
ax.set_title("Average number of motifs per peak")
vec = dfi_subset[dfi_subset.example_interval_from_task == task].pattern_name.value_counts().reindex(pattern_names, fill_value=0)
vec = vec / peaks_per_task[task]
vec.plot.bar(width=.9, ax=ax)
ax.set_ylabel(task, rotation=0, labelpad=10, ha='right')
ax.set_xlabel("Motif")
fig.savefig(fdir / '../n-instances-per-motif.per-task.avg-motifs-per-peak.pdf')
In [16]:
np.sum(dfi_subset.groupby('example_idx').size() >= 3)
Out[16]:
In [19]:
np.sum(dfi_subset.groupby('example_idx').size() >= 3) / 150908
Out[19]:
In [18]:
np.sum(dfi_subset.groupby('example_idx').size() >= 5)
Out[18]:
In [17]:
72696 / 150000
Out[17]:
In [14]:
dfi_subset.groupby('example_idx').size()
Out[14]:
In [47]:
patterns = [p for p in mr.get_all_patterns()]
In [93]:
ic = [p.seq_info_content for p in mr.get_all_patterns() if mr.n_seqlets(p.name) > 100]
In [99]:
n_seqlets = [mr.n_seqlets(p.name) for p in mr.get_all_patterns()]
In [94]:
len(ic)
Out[94]:
In [97]:
ic = [p.seq_info_content for p in mr.get_all_patterns()]
In [98]:
fig, ax = plt.subplots(figsize=get_figsize(0.4))
plt.hist(ic, bins=np.arange(0, 110, step=8), rwidth=0.9);
# plt.hist(ic, bins=15, rwidth=0.9, align='left');
plt.xlabel("IC")
plt.ylabel("Motifs")
fig.savefig(fdir / '../seq-IC.hist.all.pdf')
In [48]:
p = patterns[0]
In [50]:
p.seq_info_content
Out[50]:
Check if they occur at the same location¶
Export the files¶
In [32]:
tracks_dir = Path(ddir) / 'www/paper/data/tracks'
!mkdir -p {tracks_dir}
In [44]:
bed_cols = ['Chromosome', 'Start', 'End', 'pattern_name', 'match_weighted_p', 'strand', 'imp_weighted_p', 'seq_match_p']
duplicates = dfi_subset[bed_cols[:4]].duplicated() # Drop duplicates
dfi_subset[~duplicates][bed_cols].sort_values(bed_cols[:2]).to_csv(tracks_dir / "motif-instances.bed", sep='\t', header=False, index=False)
In [45]:
!head {tracks_dir}/motif-instances.bed
In [46]:
!wc -l {tracks_dir}/motif-instances.bed
In [49]:
# write out the region
ranges = isf.get_ranges()
ranges[['chrom', 'start', 'end', 'interval_from_task']].to_csv(tracks_dir / "regions.bed", sep='\t', header=False, index=False)
In [54]:
!wc -l {tracks_dir}/regions.bed
In [57]:
ls {model_dir}/
In [70]:
model_dir
Out[70]:
In [69]:
ls {model_dir}/log