In [1]:
# Filepaths and Hard-coded Defaults

proj_root = "/home/users/kcochran/oak/kcochran/procap_models/"
sequence_path = proj_root + "genomes/GRCh38_no_alt_analysis_set_GCA_000001405.15.fasta"
chrom_sizes = proj_root + "genomes/hg38.chrom.sizes.withrRNA"

in_window = 2114
out_window = 1000
In [2]:
# stuff to get from config file

with open("2022-07-26_17-52-41_run1_modisco_config_K562_counts.txt") as config_f:
    config_dict = {line.split()[0] : line.strip().split()[1] for line in config_f}

modisco_out_path = config_dict["modisco_out_path"]
scoring_type = config_dict["scoring_type"]
score_center_size = int(config_dict["score_center_size"])
profile_display_center_size = int(config_dict["profile_display_center_size"])
train_val_type = config_dict["train_val_type"]

# digest what's in config file

if not modisco_out_path.endswith("/"):
    modisco_out_path = modisco_out_path + "/"

assay_type, model_type, cell, accession, modisco_dir_base = modisco_out_path.split("/")[-6:-1]
ts_part1, ts_part2, run_str, _ = modisco_dir_base.split("_")
timestamp = ts_part1 + "_" + ts_part2
run = int(run_str.replace("run", ""))
In [3]:
print(modisco_out_path)
print("cell_type:", cell, accession)
print("timestamp:", timestamp)
print("run:", run)
print("scoring_type:", scoring_type)
print("score_center_size:", score_center_size)
print("profile_display_center_size:", profile_display_center_size)

/home/users/kcochran/oak/kcochran/procap_models/modisco_out/procap_bias/bpnetlite_basic_v2/K562/ENCSR261KBX/2022-07-26_17-52-41_run1_modisco/
cell_type: K562 ENCSR261KBX
timestamp: 2022-07-26_17-52-41
run: 1
scoring_type: counts
score_center_size: 1000
profile_display_center_size: 400
In [4]:
data_dir = proj_root + "/data/procap/processed/" + cell + "/" + accession + "/"
plus_bw_path = data_dir + "final.5prime.pos.bigWig"
minus_bw_path = data_dir + "final.5prime.neg.bigWig"
val_peak_path = data_dir + "peaks_uni_and_bi_" + train_val_type + ".bed.gz"

val_save_dir = proj_root + "model_out/" + assay_type + "/" + model_type + "/" + cell + "/" + accession + "/"
val_save_path = val_save_dir + timestamp + "_run" + str(run) + "_" + train_val_type

attr_save_path = val_save_dir.replace("model_out", "deepshap_out") + timestamp + "_run" + str(run) + "_deepshap"
In [5]:
# task-specific filepaths

import os

assert scoring_type in ["profile", "counts"], scoring_type

if scoring_type == "profile":
    score_type_short = "prof"
else:
    score_type_short = "count"

scores_path = attr_save_path + "_" + score_type_short + ".npy"
onehot_scores_path = attr_save_path + "_" + score_type_short + "_onehot.npy"
modisco_obj_path = modisco_out_path + "results_allChroms_" + score_type_short + "_slice" + str(score_center_size) + ".hdf5"
seqlet_path = modisco_out_path + "seqlets_" + score_type_short + ".txt"
    
tomtom_dir = modisco_out_path + "tomtom_" + score_type_short
    
assert(os.path.exists(scores_path)), scores_path
assert(os.path.exists(onehot_scores_path)), onehot_scores_path
In [6]:
# Imports, Plotting Defaults

import matplotlib.pyplot as plt
import matplotlib.font_manager as font_manager

plot_params = {
    "figure.titlesize": 22,
    "axes.titlesize": 22,
    "axes.labelsize": 20,
    "legend.fontsize": 18,
    "xtick.labelsize": 16,
    "ytick.labelsize": 16,
    "font.weight": "bold"
}
plt.rcParams.update(plot_params)

from IPython.display import display
import tqdm
tqdm.tqdm_notebook()

import numpy as np

from view_modisco_results_utils import *
from tomtom_utils import *  

/home/users/kcochran/miniconda3/envs/procap/lib/python3.7/site-packages/ipykernel_launcher.py:19: TqdmDeprecationWarning: This function will be removed in tqdm==5.0.0
Please use `tqdm.notebook.tqdm` instead of `tqdm.tqdm_notebook`
In [7]:
# Load in True Profiles and Sequences

import sys
sys.path.append('../1_train_models')

from data_loading import extract_peaks

one_hot_seqs, true_profs = extract_peaks(sequence_path, 
    plus_bw_path, minus_bw_path, val_peak_path, in_window, out_window,
    max_jitter=0, verbose=True)

one_hot_seqs = one_hot_seqs.swapaxes(1,2)
one_hot_seqs = one_hot_seqs[:, (in_window // 2 - score_center_size // 2):(in_window // 2 + score_center_size // 2), :]

Reading FASTA: 100%|██████████| 24/24 [00:12<00:00,  1.86it/s]
Loading Peaks: 3834it [00:04, 821.80it/s]
In [8]:
# Load in Coordinates of Examples
    
coords = load_coords(val_peak_path, in_window)
In [9]:
# Import SHAP scores, predicted profiles

hyp_scores = np.load(scores_path).swapaxes(1,2)
hyp_scores = hyp_scores[:, (in_window // 2 - score_center_size // 2):(in_window // 2 + score_center_size // 2), :]
pred_profs = np.exp(np.load(val_save_path + ".profs.npy"))
In [10]:
# Load modisco results object
    
tfm_obj = import_tfmodisco_results(modisco_obj_path, hyp_scores, one_hot_seqs)
In [11]:
motif_pfms, motif_hcwms, motif_cwms, \
    motif_pfms_short, num_seqlets, \
    motif_seqlets, num_metaclusters = plot_all_metaclusters(tfm_obj, one_hot_seqs, hyp_scores,
                                                            true_profs, pred_profs, coords,
                                                            in_window, out_window,
                                                            score_center_size,
                                                            profile_display_center_size)

Metacluster 1/1

Pattern 1/20

2961 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 2/20

2918 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 3/20

2206 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 4/20

1847 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 5/20

854 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 6/20

366 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 7/20

210 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 8/20

139 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 9/20

138 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 10/20

125 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 11/20

113 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 12/20

75 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 13/20

74 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 14/20

73 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 15/20

68 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 16/20

64 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 17/20

60 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 18/20

54 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 19/20

46 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)

Pattern 20/20

41 seqlets

Sequence (PFM)
Hypothetical contributions (hCWM)
Actual contributions (CWM)
In [12]:
run_and_plot_tomtom(modisco_out_path, motif_pfms, motif_hcwms, motif_pfms_short, num_metaclusters, tomtom_dir)

Metacluster 1/1

Motif 1/20

Motif IDq-valPWM
MA1513.10.0170675
MA1959.10.0245984
MA0685.20.0245984
MA1890.10.0245984
MA1713.10.0245984

Motif 2/20

Motif IDq-valPWM
MA2022.11.03823e-08
MA1833.12.79993e-08
MA1239.12.79993e-08
MA1228.16.6697e-08
MA1257.16.71537e-08

Motif 3/20

Motif IDq-valPWM
MA1615.10.00918128
MA1712.10.0298034
MA0163.10.0298034
MA1841.10.0298034
MA1976.10.0298034

Motif 4/20

Motif IDq-valPWM
MA1821.15.66981e-06
MA1832.18.78758e-06
MA1818.11.33304e-05
MA1833.13.22322e-05
MA1817.13.22322e-05

Motif 5/20

Motif IDq-valPWM
MA0374.10.0280737
MA1513.10.0280737
MA1880.10.0280737
MA1650.10.0362984
MA1961.10.0362984

Motif 6/20

Motif IDq-valPWM
MA1976.10.00290309
MA1430.10.0160478
MA1410.10.0167802
MA0163.10.0206275
MA1841.10.0341084

Motif 7/20

Motif IDq-valPWM
MA2022.17.34368e-05
MA1833.10.000415367
MA1246.10.000415367
MA1890.10.000440045
MA1257.10.00112716

Motif 8/20

Motif IDq-valPWM
MA1035.10.195739
MA1096.10.195739
MA1067.10.195739
MA1841.10.229984
MA1054.10.288355

Motif 9/20

No TOMTOM matches passing threshold

Motif 10/20

Motif IDq-valPWM
MA1880.10.0185105
MA1899.10.104708
MA1878.10.104708
MA1102.20.104708
MA0146.20.104708

Motif 11/20

Motif IDq-valPWM
MA1880.10.00256607
MA1650.10.0342193
MA1892.10.0342193
MA1051.10.0649158
MA0147.30.0649158

Motif 12/20

Motif IDq-valPWM
MA1596.10.000228822
MA1587.10.0148212
MA1232.10.0299065
MA2022.10.051092
MA1239.10.051092

Motif 13/20

No TOMTOM matches passing threshold

Motif 14/20

Motif IDq-valPWM
MA1841.10.0597562
MA1031.10.0598503
MA1615.10.0682328
MA1410.10.0682328
MA1428.10.0682328

Motif 15/20

Motif IDq-valPWM
MA1880.10.0373465
MA1713.10.0448176
MA0374.10.0608831
MA1961.10.075625
MA0535.10.075625

Motif 16/20

Motif IDq-valPWM
MA1540.20.0347133
MA1990.10.11541

Motif 17/20

Motif IDq-valPWM
MA0599.10.295108
MA0516.30.295108
MA0066.10.295108
MA0742.20.295108
MA1833.10.295108

Motif 18/20

No TOMTOM matches passing threshold

Motif 19/20

Motif IDq-valPWM
MA1587.10.00398675
MA1596.10.00398675
MA1581.10.0429213
MA0358.10.375661
MA0003.40.375661

Motif 20/20

Motif IDq-valPWM
MA1650.10.0274155
MA1053.10.0274155
MA0375.10.0274155
MA1833.10.0274155
MA1051.10.0274155