Link to results

DiChIPMunk: on peaks and on multi-task seqlets and on single-task seqlets

HOMER: on peaks and on multi-task seqlets and on single-task seqlets

MEME: on peaks and on multitask seqlets and on single-task seqlets

import sys
import os
sys.path.append(os.path.abspath("/users/amtseng/tfmodisco/src/"))
from util import figure_to_vdom_image
import motif.read_motifs as read_motifs
from motif.read_motifs import pfm_to_pwm
import plot.viz_sequence as viz_sequence
import numpy as np
import matplotlib.pyplot as plt
import vdom.helpers as vdomh
from IPython.display import display

# Define parameters/fetch arguments
tf_name = os.environ["TFM_TF_NAME"]
multitask_fold = int(os.environ["TFM_MULTITASK_FOLD"])

if "TFM_TASK_INDEX" in os.environ:
    task_index = int(os.environ["TFM_TASK_INDEX"])
    singletask_fold = int(os.environ["TFM_SINGLETASK_FOLD"])
else:
    task_index = None
    singletask_fold = None
    
print("TF name: %s" % tf_name)
print("Multi-task fold: %s" % multitask_fold)
print("Task index: %s" % task_index)
print("Single-task fold: %s" % singletask_fold)

TF name: JUND
Multi-task fold: 7
Task index: 0
Single-task fold: 7

# Define paths and constants
base_path = "/users/amtseng/tfmodisco/results/classic_motifs/"

multitask_seqlets_dir = os.path.join(
    base_path, "seqlets", "multitask_profile_finetune",
    "%s_multitask_profile_finetune_fold%s" % (tf_name, multitask_fold)
)

if task_index is None:
    peaks_path = os.path.join(base_path, "peaks", tf_name, "%s_peaks_taskall" % tf_name)
    multitask_profile_seqlets_path = os.path.join(
        multitask_seqlets_dir,
        "%s_seqlets_profile_taskall" % tf_name
    )
    multitask_count_seqlets_path = os.path.join(
        multitask_seqlets_dir,
        "%s_seqlets_count_taskall" % tf_name
    )
else:
    peaks_path = os.path.join(base_path, "peaks", tf_name, "%s_peaks_task%d" % (tf_name, task_index))
    multitask_profile_seqlets_path = os.path.join(
        multitask_seqlets_dir,
        "%s_seqlets_profile_task%d" % (tf_name, task_index)
    )
    multitask_count_seqlets_path = os.path.join(
        multitask_seqlets_dir,
        "%s_seqlets_count_task%d" % (tf_name, task_index)
    )
    
    singletask_seqlets_dir = os.path.join(
        base_path, "seqlets", "singletask_profile_finetune",
        "%s_singletask_profile_finetune_fold%s" % (tf_name, singletask_fold),
        "task_%d" % task_index
    )
    singletask_profile_seqlets_path = os.path.join(
        singletask_seqlets_dir,
        "%s_seqlets_profile_task%d" % (tf_name, task_index)
    )
    singletask_count_seqlets_path = os.path.join(
        singletask_seqlets_dir,
        "%s_seqlets_count_task%d" % (tf_name, task_index)
    )

Helper functions

def show_peaks_motif_table(results_path, mode):
    """
    Shows a table of motifs from the given results path.
    `mode` is either `dichipmunk`, `homer`, `meme`, or `memechip`.
    """
    assert mode in ("dichipmunk", "homer", "meme", "memechip")
    if mode == "dichipmunk":
        score_name = "Supporting sequences"
        pfms, score_vals = read_motifs.import_dichipmunk_pfms(results_path)
    elif mode == "homer":
        score_name = "Log enrichment"
        pfms, score_vals = read_motifs.import_homer_pfms(results_path)
    elif mode == "meme":
        score_name = "E-value"
        pfms, score_vals = read_motifs.import_meme_pfms(results_path)
    else:
        score_name = "E-value"
        pfms, score_vals = read_motifs.import_meme_pfms(
            os.path.join(results_path, "meme_out")
        )
        
    colgroup = vdomh.colgroup(
        vdomh.col(style={"width": "5%"}),
        vdomh.col(style={"width": "5%"}),
        vdomh.col(style={"width": "40%"})
    )
    header = vdomh.thead(
        vdomh.tr(
            vdomh.th("Motif", style={"text-align": "center"}),
            vdomh.th(score_name, style={"text-align": "center"}),
            vdomh.th("PWM", style={"text-align": "center"})
        )
    )

    body = []
    for i, pfm in enumerate(pfms):
        pwm = pfm_to_pwm(pfm)
        if np.sum(pwm[:, [0, 2]]) < 0.5 * np.sum(pwm):
            # Flip to purine-rich version
            pwm = np.flip(pwm, axis=(0, 1))
        fig = viz_sequence.plot_weights(pwm, figsize=(20, 4), return_fig=True)
        fig.tight_layout()

        body.append(
            vdomh.tr(
                vdomh.td(str(i + 1)),
                vdomh.td(str(score_vals[i])),
                vdomh.td(figure_to_vdom_image(fig))
            )
        )

    display(vdomh.table(colgroup, header, vdomh.tbody(*body)))
    plt.close("all")

def show_seqlets_motif_table(profile_results_path, count_results_path, mode):
    """
    Shows a table of motifs from the given results path.
    `mode` is either `dichipmunk`, `homer`, `meme`, or `memechip`
    """
    assert mode in ("dichipmunk", "homer", "meme", "memechip")
    if mode == "dichipmunk":
        score_name = "Supporting sequences"
        p_pfms, p_score_vals = read_motifs.import_dichipmunk_pfms(profile_results_path)
        c_pfms, c_score_vals = read_motifs.import_dichipmunk_pfms(count_results_path)
    elif mode == "homer":
        score_name = "Log enrichment"
        p_pfms, p_score_vals = read_motifs.import_homer_pfms(profile_results_path)
        c_pfms, c_score_vals = read_motifs.import_homer_pfms(count_results_path)
    elif mode == "meme":
        score_name = "E-value"
        p_pfms, p_score_vals = read_motifs.import_meme_pfms(profile_results_path)
        c_pfms, c_score_vals = read_motifs.import_meme_pfms(count_results_path)
    else:
        score_name = "E-value"
        p_pfms, p_score_vals = read_motifs.import_meme_pfms(
            os.path.join(profile_results_path, "meme_out")
        )
        c_pfms, c_score_vals = read_motifs.import_meme_pfms(
            os.path.join(count_results_path, "meme_out")
        )
        
    colgroup = vdomh.colgroup(
        vdomh.col(style={"width": "5%"}),
        vdomh.col(style={"width": "5%"}),
        vdomh.col(style={"width": "40%"}),
        vdomh.col(style={"width": "5%"}),
        vdomh.col(style={"width": "40%"})
    )
    header = vdomh.thead(
        vdomh.tr(
            vdomh.th("Motif", style={"text-align": "center"}),
            vdomh.th(score_name + " (profile)", style={"text-align": "center"}),
            vdomh.th("PWM (profile)", style={"text-align": "center"}),
            vdomh.th(score_name + " (count)", style={"text-align": "center"}),
            vdomh.th("PWM (count)", style={"text-align": "center"})
        )
    )

    body = []
    for i in range(max(len(p_pfms), len(c_pfms))):
        rows = [vdomh.td(str(i + 1))]
        if i < len(p_pfms):
            pwm = pfm_to_pwm(p_pfms[i])
            if np.sum(pwm[:, [0, 2]]) < 0.5 * np.sum(pwm):
                # Flip to purine-rich version
                pwm = np.flip(pwm, axis=(0, 1))
            fig = viz_sequence.plot_weights(pwm, figsize=(20, 4), return_fig=True)
            fig.tight_layout()
            rows.extend([
                vdomh.td(str(p_score_vals[i])),
                vdomh.td(figure_to_vdom_image(fig))
            ])
        else:
            rows.extend([vdomh.td(), vdomh.td()])
            
        if i < len(c_pfms):
            pwm = pfm_to_pwm(c_pfms[i])
            if np.sum(pwm[:, [0, 2]]) < 0.5 * np.sum(pwm):
                # Flip to purine-rich version
                pwm = np.flip(pwm, axis=(0, 1))
            fig = viz_sequence.plot_weights(pwm, figsize=(20, 4), return_fig=True)
            fig.tight_layout()
            rows.extend([
                vdomh.td(str(c_score_vals[i])),
                vdomh.td(figure_to_vdom_image(fig))
            ])
        else:
            rows.extend([vdomh.td(), vdomh.td()])
            

        body.append(vdomh.tr(*rows))

    display(vdomh.table(colgroup, header, vdomh.tbody(*body)))
    plt.close("all")

Show benchmark motifs

DiChIPMunk on peaks

show_peaks_motif_table(os.path.join(peaks_path, "dichipmunk"), "dichipmunk")

Motif	Supporting sequences	PWM
1	2000
2	2000
3	2000
4	1975
5	1237
6	684
7	400
8	255
9	115
10	111

DiChIPMunk on multi-task seqlets

show_seqlets_motif_table(
    os.path.join(multitask_profile_seqlets_path, "dichipmunk"),
    os.path.join(multitask_count_seqlets_path, "dichipmunk"),
    "dichipmunk"
)

Motif	Supporting sequences (profile)	PWM (profile)	Supporting sequences (count)	PWM (count)
1	10171		9204
2	3541		2244
3	1528		397
4	624		353
5			146
6			6

DiChIPMunk on single-task seqlets

if task_index is not None:
    show_seqlets_motif_table(
        os.path.join(singletask_profile_seqlets_path, "dichipmunk"),
        os.path.join(singletask_count_seqlets_path, "dichipmunk"),
        "dichipmunk"
    )

Motif	Supporting sequences (profile)	PWM (profile)	Supporting sequences (count)	PWM (count)
1	8521		10490
2	4126		4534
3	856		566
4	440		136
5	7		141
6			57
7			31
8			1

HOMER on peaks

show_peaks_motif_table(os.path.join(peaks_path, "homer"), "homer")

Motif	Log enrichment	PWM
1	-16753.851436
2	-4073.382566
3	-4010.093235
4	-3679.353153
5	-3629.883251
6	-3019.06886
7	-2383.103977
8	-2062.329642
9	-1883.377548
10	-1207.809743
11	-1075.856862
12	-1017.902963
13	-985.699754
14	-907.58392
15	-710.688609
16	-626.610637
17	-602.7199
18	-513.152551
19	-493.260766
20	-178.322196

HOMER on multi-task seqlets

show_seqlets_motif_table(
    os.path.join(multitask_profile_seqlets_path, "homer"),
    os.path.join(multitask_count_seqlets_path, "homer"),
    "homer"
)

/users/amtseng/tfmodisco/src/plot/viz_sequence.py:152: RuntimeWarning: More than 20 figures have been opened. Figures created through the pyplot interface (`matplotlib.pyplot.figure`) are retained until explicitly closed and may consume too much memory. (To control this warning, see the rcParam `figure.max_open_warning`).
  fig = plt.figure(figsize=figsize)

Motif	Log enrichment (profile)	PWM (profile)	Log enrichment (count)	PWM (count)
1	-23919.701228		-38063.422152
2	-2504.832579		-3176.305662
3	-806.397651		-412.290496
4	-609.158043		-289.756436
5	-447.891197		-262.867292
6	-389.221737		-252.226006
7	-387.048369		-189.911639
8	-350.308822		-140.968486
9	-316.867704		-111.550292
10	-311.201783		-107.250385
11	-296.571217		-106.181037
12	-224.648009		-91.371992
13	-214.146019		-84.61473
14	-133.304097		-80.204124
15	-119.479438		-53.650211
16	-118.889463		-33.263825
17	-81.666082		-26.977346
18	-25.81566

HOMER on single-task seqlets

if task_index is not None:
    show_seqlets_motif_table(
        os.path.join(singletask_profile_seqlets_path, "homer"),
        os.path.join(singletask_count_seqlets_path, "homer"),
        "homer"
    )

Motif	Log enrichment (profile)	PWM (profile)	Log enrichment (count)	PWM (count)
1	-15690.096072		-27388.952787
2	-4417.203786		-5778.26107
3	-2852.863338		-4221.676602
4	-1471.745233		-1660.357363
5	-1244.508496		-602.387024
6	-887.123245		-599.550135
7	-802.587454		-524.067346
8	-768.010763		-491.698005
9	-618.577642		-470.779283
10	-364.370175		-407.775526
11	-323.987148		-349.877731
12	-309.247978		-175.355576
13	-215.524859		-64.072057
14	-189.48593
15	-179.047449
16	-176.568854
17	-153.972922
18	-115.126555
19	-40.149588

MEMEChIP on peaks

show_peaks_motif_table(os.path.join(peaks_path, "memechip"), "memechip")

Motif	E-value	PWM
1	0.0
2	1.1e-82
3	1.1e-35
4	2.7e-21
5	3.3e-21
6	5.3e-21
7	6.5e-17
8	5.2e-12
9	4.4e-10
10	2.5e-07

MEME on multi-task seqlets

show_seqlets_motif_table(
    os.path.join(multitask_profile_seqlets_path, "meme"),
    os.path.join(multitask_count_seqlets_path, "meme"),
    "meme"
)

Motif	E-value (profile)	PWM (profile)	E-value (count)	PWM (count)
1	0.0		0.0
2	1.2e-93		5.2e-144
3	1.8e-38		3.6e-05
4	0.00035		0.21
5	0.0014		1000.0
6	12000.0		8400000.0
7	7600.0		5800000.0
8	17000.0		12000000.0
9	38000.0		14000000.0
10	5500000.0		14000000.0

MEME on single-task seqlets

if task_index is not None:
    show_seqlets_motif_table(
        os.path.join(singletask_profile_seqlets_path, "meme"),
        os.path.join(singletask_count_seqlets_path, "meme"),
        "meme"
    )

Motif	E-value (profile)	PWM (profile)	E-value (count)	PWM (count)
1	0.0		0.0
2	4.1e-81		3.6e-199
3	5.5e-39		1.7e-95
4	3e-27		4e-22
5	1.2e-25		0.037
6	1.1e-16		0.24
7	0.00085		570.0
8	0.023		780.0
9	5500.0		1100.0
10	7600.0		10000.0