TF-MoDISco results

Summary of motifs

TOMTOM matches to motifs

Sample of seqlets for each motif

In [1]:
import os
import util
from tomtom import match_motifs_to_database
import viz_sequence
import numpy as np
import pandas as pd
import sklearn.cluster
import scipy.cluster.hierarchy
import matplotlib.pyplot as plt
import matplotlib.font_manager as font_manager
import vdom.helpers as vdomh
from IPython.display import display
import tqdm
tqdm.tqdm_notebook()

/users/amtseng/miniconda3/envs/tfmodisco-mini/lib/python3.7/site-packages/ipykernel_launcher.py:14: TqdmDeprecationWarning: This function will be removed in tqdm==5.0.0
Please use `tqdm.notebook.tqdm` instead of `tqdm.tqdm_notebook`
Out[1]:
0it [00:00, ?it/s]
In [2]:
# Plotting defaults
font_manager.fontManager.ttflist.extend(
    font_manager.createFontList(
        font_manager.findSystemFonts(fontpaths="/users/amtseng/modules/fonts")
    )
)
plot_params = {
    "figure.titlesize": 22,
    "axes.titlesize": 22,
    "axes.labelsize": 20,
    "legend.fontsize": 18,
    "xtick.labelsize": 16,
    "ytick.labelsize": 16,
    "font.family": "Roboto",
    "font.weight": "bold"
}
plt.rcParams.update(plot_params)

/users/amtseng/miniconda3/envs/tfmodisco-mini/lib/python3.7/site-packages/ipykernel_launcher.py:4: MatplotlibDeprecationWarning: 
The createFontList function was deprecated in Matplotlib 3.2 and will be removed two minor releases later. Use FontManager.addfont instead.
  after removing the cwd from sys.path.

Define constants and paths

In [3]:
# Define parameters/fetch arguments
preds_path = os.environ["TFM_PRED_PATH"]
shap_scores_path = os.environ["TFM_SHAP_PATH"]
tfm_results_path = os.environ["TFM_TFM_PATH"]
peak_bed_paths = [os.environ["TFM_PEAKS_PATH"]]

print("Predictions path: %s" % preds_path)
print("DeepSHAP scores path: %s" % shap_scores_path)
print("TF-MoDISco results path: %s" % tfm_results_path)
print("Peaks path: %s" % peak_bed_paths[0])

Predictions path: /users/zahoor/TF-Atlas/02-16-2021/predictions/ENCSR240PRQ/profile_predictions.h5
DeepSHAP scores path: /users/zahoor/TF-Atlas/02-16-2021/shap/ENCSR240PRQ/profile_scores_alex_format.h5
TF-MoDISco results path: /users/zahoor/TF-Atlas/02-16-2021/modisco/ENCSR240PRQ/profile/modisco_results.hd5
Peaks path: /users/zahoor/TF-Atlas/data/idr_peaks//ENCFF463FGL.bed.gz
In [4]:
# Define constants
input_length, profile_length = 2114, 1000
shap_score_center_size = 400
profile_display_center_size = 400
hyp_score_key = "hyp_scores"
task_index = None

Helper functions

In [5]:
def extract_profiles_and_coords(
    seqlets_arr, one_hot_seqs, hyp_scores, true_profs, pred_profs, pred_coords,
    input_length, profile_length, input_center_cut_size, profile_center_cut_size,
    task_index=None
):
    """
    From the seqlets object of a TF-MoDISco pattern's seqlets and alignments,
    extracts the predicted and observed profiles of the model, as well as the
    set of coordinates for the seqlets.
    Arguments:
        `seqlets_arr`: a TF-MoDISco pattern's seqlets object array (N-array)
        `one_hot_seqs`: an N x R x 4 array of input sequences, where R is
            the cut centered size
        `hyp_scores`: an N x R x 4 array of hypothetical importance scores
        `true_profs`: an N x T x O x 2 array of true profile counts
        `pred_profs`: an N x T x O x 2 array of predicted profile probabilities
        `pred_coords`: an N x 3 object array of coordinates for the input sequence
            underlying the predictions
        `input_length`: length of original input sequences, I
        `profile_length`: length of profile predictions, O
        `input_center_cut_size`: centered cut size of SHAP scores used
        `profile_center_cut_size`: size to cut profiles to when returning them, P
        `task_index`: index of task to focus on for profiles; if None, returns
            profiles for all tasks
    Returns an N x (T or 1) x P x 2 array of true profile counts, an
    N x (T or 1) x P x 2 array of predicted profile probabilities, an N x Q x 4
    array of one-hot seqlet sequences, an N x Q x 4 array of hypothetical seqlet
    importance scores, and an N x 3 object array of seqlet coordinates, where P
    is the profile cut size and Q is the seqlet length. Returned profiles are
    centered at the same center as the seqlets.
    Note that it is important that the seqlet indices match exactly with the indices
    out of the N. This should be the exact sequences in the original SHAP scores.
    """
    true_seqlet_profs, pred_seqlet_profs, seqlet_seqs, seqlet_hyps, seqlet_coords = [], [], [], [], []
    
    def seqlet_coord_to_profile_coord(seqlet_coord):
        return seqlet_coord + ((input_length - input_center_cut_size) // 2) - ((input_length - profile_length) // 2)
    
    def seqlet_coord_to_input_coord(seqlet_coord):
        return seqlet_coord + ((input_length - input_center_cut_size) // 2)
        
    # For each seqlet, fetch the true/predicted profiles
    for seqlet in seqlets_arr:
        coord_index = seqlet.coor.example_idx
        seqlet_start = seqlet.coor.start
        seqlet_end = seqlet.coor.end
        seqlet_rc = seqlet.coor.is_revcomp
        
        # Get indices of profile to cut out
        seqlet_center = (seqlet_start + seqlet_end) // 2
        prof_center = seqlet_coord_to_profile_coord(seqlet_center)
        prof_start = prof_center - (profile_center_cut_size // 2)
        prof_end = prof_start + profile_center_cut_size
        
        if task_index is None or true_profs.shape[1] == 1:
            # Use all tasks if the predictions only have 1 task to begin with
            task_start, task_end = None, None
        else:
            task_start, task_end = task_index, task_index + 1
            
        true_prof = true_profs[coord_index, task_start:task_end, prof_start:prof_end]  # (T or 1) x P x 2
        pred_prof = pred_profs[coord_index, task_start:task_end, prof_start:prof_end]  # (T or 1) x P x 2
        
        true_seqlet_profs.append(true_prof)
        pred_seqlet_profs.append(pred_prof)
        
        # The one-hot-sequences and hypothetical scores are assumed to already by cut/centered,
        # so the indices match the seqlet indices
        if seqlet_rc:
            seqlet_seqs.append(np.flip(one_hot_seqs[coord_index, seqlet_start:seqlet_end], axis=(0, 1)))
            seqlet_hyps.append(np.flip(hyp_scores[coord_index, seqlet_start:seqlet_end], axis=(0, 1)))
        else:
            seqlet_seqs.append(one_hot_seqs[coord_index, seqlet_start:seqlet_end])
            seqlet_hyps.append(hyp_scores[coord_index, seqlet_start:seqlet_end])
            
        # Get the coordinates of the seqlet based on the input coordinates
        inp_start = seqlet_coord_to_input_coord(seqlet_start)
        inp_end = seqlet_coord_to_input_coord(seqlet_end)
        chrom, start, _ = pred_coords[coord_index]
        seqlet_coords.append([chrom, start + inp_start, start + inp_end])
    
    return np.stack(true_seqlet_profs), np.stack(pred_seqlet_profs), np.stack(seqlet_seqs), np.stack(seqlet_hyps), np.array(seqlet_coords, dtype=object)
In [6]:
def plot_profiles(seqlet_true_profs, seqlet_pred_profs, kmeans_clusters=5):
    """
    Plots the given profiles with a heatmap.
    Arguments:
        `seqlet_true_profs`: an N x O x 2 NumPy array of true profiles, either as raw
            counts or probabilities (they will be normalized)
        `seqlet_pred_profs`: an N x O x 2 NumPy array of predicted profiles, either as
            raw counts or probabilities (they will be normalized)
        `kmeans_cluster`: when displaying profile heatmaps, there will be this
            many clusters
    """
    assert len(seqlet_true_profs.shape) == 3
    assert seqlet_true_profs.shape == seqlet_pred_profs.shape
    num_profs, width, _ = seqlet_true_profs.shape

    # First, normalize the profiles along the output profile dimension
    def normalize(arr, axis=0):
        arr_sum = np.sum(arr, axis=axis, keepdims=True)
        arr_sum[arr_sum == 0] = 1  # If 0, keep 0 as the quotient instead of dividing by 0
        return arr / arr_sum
    true_profs_norm = normalize(seqlet_true_profs, axis=1)
    pred_profs_norm = normalize(seqlet_pred_profs, axis=1)

    # Compute the mean profiles across all examples
    true_profs_mean = np.mean(true_profs_norm, axis=0)
    pred_profs_mean = np.mean(pred_profs_norm, axis=0)

    # Perform k-means clustering on the predicted profiles, with the strands pooled
    kmeans_clusters = max(5, num_profs // 50)  # Set number of clusters based on number of profiles, with minimum
    kmeans = sklearn.cluster.KMeans(n_clusters=kmeans_clusters)
    cluster_assignments = kmeans.fit_predict(
        np.reshape(pred_profs_norm, (pred_profs_norm.shape[0], -1))
    )

    # Perform hierarchical clustering on the cluster centers to determine optimal ordering
    kmeans_centers = kmeans.cluster_centers_
    cluster_order = scipy.cluster.hierarchy.leaves_list(
        scipy.cluster.hierarchy.optimal_leaf_ordering(
            scipy.cluster.hierarchy.linkage(kmeans_centers, method="centroid"), kmeans_centers
        )
    )

    # Order the profiles so that the cluster assignments follow the optimal ordering
    cluster_inds = []
    for cluster_id in cluster_order:
        cluster_inds.append(np.where(cluster_assignments == cluster_id)[0])
    cluster_inds = np.concatenate(cluster_inds)

    # Compute a matrix of profiles, normalized to the maximum height, ordered by clusters
    def make_profile_matrix(flat_profs, order_inds):
        matrix = flat_profs[order_inds]
        maxes = np.max(matrix, axis=1, keepdims=True)
        maxes[maxes == 0] = 1  # If 0, keep 0 as the quotient instead of dividing by 0
        return matrix / maxes
    true_matrix = make_profile_matrix(true_profs_norm, cluster_inds)
    pred_matrix = make_profile_matrix(pred_profs_norm, cluster_inds)

    # Create a figure with the right dimensions
    mean_height = 4
    heatmap_height = min(num_profs * 0.004, 8)
    fig_height = mean_height + (2 * heatmap_height)
    fig, ax = plt.subplots(
        3, 2, figsize=(16, fig_height), sharex=True,
        gridspec_kw={
            "width_ratios": [1, 1],
            "height_ratios": [mean_height / fig_height, heatmap_height / fig_height, heatmap_height / fig_height]
        }
    )

    # Plot the average predictions
    ax[0, 0].plot(true_profs_mean[:, 0], color="darkslateblue")
    ax[0, 0].plot(-true_profs_mean[:, 1], color="darkorange")
    ax[0, 1].plot(pred_profs_mean[:, 0], color="darkslateblue")
    ax[0, 1].plot(-pred_profs_mean[:, 1], color="darkorange")

    # Set axes on average predictions
    max_mean_val = max(np.max(true_profs_mean), np.max(pred_profs_mean))
    mean_ylim = max_mean_val * 1.05  # Make 5% higher
    ax[0, 0].set_title("True profiles")
    ax[0, 0].set_ylabel("Average probability")
    ax[0, 1].set_title("Predicted profiles")
    for j in (0, 1):
        ax[0, j].set_ylim(-mean_ylim, mean_ylim)
        ax[0, j].label_outer()

    # Plot the heatmaps
    ax[1, 0].imshow(true_matrix[:, :, 0], interpolation="nearest", aspect="auto", cmap="Blues")
    ax[1, 1].imshow(pred_matrix[:, :, 0], interpolation="nearest", aspect="auto", cmap="Blues")
    ax[2, 0].imshow(true_matrix[:, :, 1], interpolation="nearest", aspect="auto", cmap="Oranges")
    ax[2, 1].imshow(pred_matrix[:, :, 1], interpolation="nearest", aspect="auto", cmap="Oranges")

    # Set axes on heatmaps
    for i in (1, 2):
        for j in (0, 1):
            ax[i, j].set_yticks([])
            ax[i, j].set_yticklabels([])
            ax[i, j].label_outer()
    width = true_matrix.shape[1]
    delta = 100
    num_deltas = (width // 2) // delta
    labels = list(range(max(-width // 2, -num_deltas * delta), min(width // 2, num_deltas * delta) + 1, delta))
    tick_locs = [label + max(width // 2, num_deltas * delta) for label in labels]
    for j in (0, 1):
        ax[2, j].set_xticks(tick_locs)
        ax[2, j].set_xticklabels(labels)
        ax[2, j].set_xlabel("Distance from peak summit (bp)")

    fig.tight_layout()
    plt.show()
In [7]:
def get_summit_distances(coords, peak_table):
    """
    Given a set of coordinates, computes the distance of the center of each
    coordinate to the nearest summit.
    Arguments:
        `coords`: an N x 3 object array of coordinates
        `peak_table`: a 6-column table of peak data, as imported by
            `import_peak_table`
    Returns and N-array of integers, which is the distance of each coordinate
    midpoint to the nearest coordinate.
    """
    chroms = coords[:, 0]
    midpoints = (coords[:, 1] + coords[:, 2]) // 2
    dists = []
    for i in range(len(coords)):
        chrom = chroms[i]
        midpoint = midpoints[i]
        rows = peak_table[peak_table["chrom"] == chrom]
        dist_arr = (midpoint - rows["summit"]).values
        min_dist = dist_arr[np.argmin(np.abs(dist_arr))]
        dists.append(min_dist)
    return np.array(dists)
In [8]:
def plot_summit_dists(summit_dists):
    """
    Plots the distribution of seqlet distances to summits.
    Arguments:
        `summit_dists`: the array of distances as returned by
            `get_summit_distances`
    """
    plt.figure(figsize=(8, 6))
    num_bins = max(len(summit_dists) // 30, 20)
    plt.hist(summit_dists, bins=num_bins, color="purple")
    plt.title("Histogram of distance of seqlets to peak summits")
    plt.xlabel("Signed distance from seqlet center to nearest peak summit (bp)")
    plt.show()

Import SHAP scores, profile predictions, and TF-MoDISco results

In [9]:
# Import SHAP coordinates and one-hot sequences
hyp_scores, _, one_hot_seqs, shap_coords = util.import_shap_scores(shap_scores_path, hyp_score_key, center_cut_size=shap_score_center_size, remove_non_acgt=False)
# This cuts the sequences/scores off just as how TF-MoDISco saw them, but the coordinates are uncut

Importing SHAP scores: 100%|██████████| 60/60 [00:16<00:00,  3.60it/s]
In [10]:
# Import the set of all profiles and their coordinates
true_profs, pred_profs, all_pred_coords = util.import_profiles(preds_path)



















In [11]:



    


# Import the set of peaks
peak_table = util.import_peak_table(peak_bed_paths)



    













In [12]:



    


# Subset the predicted profiles/coordinates to the task-specific SHAP coordinates/scores
shap_coords_table = pd.DataFrame(shap_coords, columns=["chrom", "start", "end"])
pred_coords_table = pd.DataFrame(all_pred_coords, columns=["chrom", "start", "end"])

subset_inds = pred_coords_table.reset_index().drop_duplicates(["chrom", "start", "end"]).merge(
    shap_coords_table.reset_index(), on=["chrom", "start", "end"]
).sort_values("index_y")["index_x"].values

true_profs = true_profs[subset_inds]
pred_profs = pred_profs[subset_inds]
pred_coords = all_pred_coords[subset_inds]

# Make sure the coordinates all match
assert np.all(pred_coords == shap_coords)



    













In [13]:



    


# Import the TF-MoDISco results object
tfm_obj = util.import_tfmodisco_results(tfm_results_path, hyp_scores, one_hot_seqs, shap_score_center_size)



    















Plot some SHAP score tracks
Plot the central region of some randomly selected actual importance scores













In [14]:



    


for index in np.random.choice(hyp_scores.shape[0], size=5, replace=False):
    viz_sequence.plot_weights((hyp_scores[index] * one_hot_seqs[index])[100:300], subticks_frequency=100)



    















    

















    

















    

















    

















    






























Plot TF-MoDISco results
Plot all motifs by metacluster













In [15]:



    


motif_pfms, motif_hcwms, motif_cwms = [], [], []  # Save the trimmed PFMs, hCWMs, and CWMs
motif_pfms_short = []  # PFMs that are even more trimmed (for TOMTOM)
num_seqlets = []  # Number of seqlets for each motif
motif_seqlets = []  # Save seqlets of each motif
metaclusters = tfm_obj.metacluster_idx_to_submetacluster_results
num_metaclusters = len(metaclusters.keys())
for metacluster_i, metacluster_key in enumerate(metaclusters.keys()):
    metacluster = metaclusters[metacluster_key]
    display(vdomh.h3("Metacluster %d/%d" % (metacluster_i + 1, num_metaclusters)))
    patterns = metacluster.seqlets_to_patterns_result.patterns
    if not patterns:
        break
    motif_pfms.append([])
    motif_hcwms.append([])
    motif_cwms.append([])
    motif_pfms_short.append([])
    num_seqlets.append([])
    motif_seqlets.append([])
    num_patterns = len(patterns)
    for pattern_i, pattern in enumerate(patterns):
        seqlets = pattern.seqlets
        display(vdomh.h4("Pattern %d/%d" % (pattern_i + 1, num_patterns)))
        display(vdomh.p("%d seqlets" % len(seqlets)))
        
        pfm = pattern["sequence"].fwd
        hcwm = pattern["task0_hypothetical_contribs"].fwd
        cwm = pattern["task0_contrib_scores"].fwd
        
        pfm_fig = viz_sequence.plot_weights(pfm, subticks_frequency=10, return_fig=True)
        hcwm_fig = viz_sequence.plot_weights(hcwm, subticks_frequency=10, return_fig=True)
        cwm_fig = viz_sequence.plot_weights(cwm, subticks_frequency=10, return_fig=True)
        pfm_fig.tight_layout()
        hcwm_fig.tight_layout()
        cwm_fig.tight_layout()
        
        motif_table = vdomh.table(
            vdomh.tr(
                vdomh.td("Sequence (PFM)"),
                vdomh.td(util.figure_to_vdom_image(pfm_fig))
            ),
            vdomh.tr(
                vdomh.td("Hypothetical contributions (hCWM)"),
                vdomh.td(util.figure_to_vdom_image(hcwm_fig))
            ),
            vdomh.tr(
                vdomh.td("Actual contributions (CWM)"),
                vdomh.td(util.figure_to_vdom_image(cwm_fig))
            )
        )
        display(motif_table)
        plt.close("all")  # Remove all standing figures
        
        # Trim motif based on total contribution
        score = np.sum(np.abs(cwm), axis=1)
        trim_thresh = np.max(score) * 0.2  # Cut off anything less than 20% of max score
        pass_inds = np.where(score >= trim_thresh)[0]
        
        short_trimmed_pfm = pfm[np.min(pass_inds): np.max(pass_inds) + 1]
        motif_pfms_short[-1].append(short_trimmed_pfm)
        
        # Expand trimming to +/- 4bp on either side
        start, end = max(0, np.min(pass_inds) - 4), min(len(cwm), np.max(pass_inds) + 4 + 1)
        trimmed_pfm = pfm[start:end]
        trimmed_hcwm = hcwm[start:end]
        trimmed_cwm = cwm[start:end]
        
        motif_pfms[-1].append(trimmed_pfm)
        motif_hcwms[-1].append(trimmed_hcwm)
        motif_cwms[-1].append(trimmed_cwm)
        
        num_seqlets[-1].append(len(seqlets))
        
        seqlet_true_profs, seqlet_pred_profs, seqlet_seqs, seqlet_hyps, seqlet_coords = extract_profiles_and_coords(
            seqlets, one_hot_seqs, hyp_scores, true_profs, pred_profs, pred_coords,
            input_length, profile_length, shap_score_center_size,
            profile_display_center_size, task_index=task_index
        )
        
        motif_seqlets[-1].append((seqlet_seqs, seqlet_hyps))

        assert np.allclose(np.sum(seqlet_seqs, axis=0) / len(seqlet_seqs), pattern["sequence"].fwd)
        # ^Sanity check: PFM derived from seqlets match the PFM stored in the pattern
        plot_profiles(
            # Flatten to NT x O x 2
            np.reshape(seqlet_true_profs, (-1, seqlet_true_profs.shape[2], seqlet_true_profs.shape[3])),
            np.reshape(seqlet_pred_profs, (-1, seqlet_pred_profs.shape[2], seqlet_pred_profs.shape[3]))
        )
        
        summit_dists = get_summit_distances(seqlet_coords, peak_table)
        plot_summit_dists(summit_dists)



    















    







Metacluster 1/2










    







Pattern 1/3










    







11154 seqlets










    







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










    

















    

















    







Pattern 2/3










    







320 seqlets










    







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










    

















    

















    







Pattern 3/3










    







41 seqlets










    







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










    

















    

















    







Metacluster 2/2










    







Pattern 1/10










    







820 seqlets










    







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










    

















    

















    







Pattern 2/10










    







562 seqlets










    







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










    

















    

















    







Pattern 3/10










    







507 seqlets










    







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










    

















    

















    







Pattern 4/10










    







172 seqlets










    







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










    

















    

















    







Pattern 5/10










    







170 seqlets










    







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










    

















    

















    







Pattern 6/10










    







101 seqlets










    







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










    

















    

















    







Pattern 7/10










    







100 seqlets










    







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










    

















    

















    







Pattern 8/10










    







59 seqlets










    







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










    

















    

















    







Pattern 9/10










    







39 seqlets










    







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










    

















    

















    







Pattern 10/10










    







36 seqlets










    







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










    

















    






























Summary of motifs
Motifs are trimmed based on information content, and presented in descending order by number of supporting seqlets. The motifs are separated by metacluster. The motifs are presented as hCWMs. The forward orientation is defined as the orientation that is richer in purines.













In [16]:



    


colgroup = vdomh.colgroup(
    vdomh.col(style={"width": "5%"}),
    vdomh.col(style={"width": "5%"}),
    vdomh.col(style={"width": "45%"}),
    vdomh.col(style={"width": "45%"})
)
header = vdomh.thead(
    vdomh.tr(
        vdomh.th("#", style={"text-align": "center"}),
        vdomh.th("Seqlets", style={"text-align": "center"}),
        vdomh.th("Forward", style={"text-align": "center"}),
        vdomh.th("Reverse", style={"text-align": "center"})
    )
)

for i in range(len(motif_hcwms)):
    display(vdomh.h3("Metacluster %d/%d" % (i + 1, num_metaclusters)))
    body = []
    for j in range(len(motif_hcwms[i])):
        motif = motif_hcwms[i][j]
        if np.sum(motif[:, [0, 2]]) > 0.5 * np.sum(motif):
            # Forward is purine-rich, reverse-complement is pyrimidine-rich
            f, rc = motif, np.flip(motif, axis=(0, 1))
        else:
            f, rc = np.flip(motif, axis=(0, 1)), motif
            
        f_fig = viz_sequence.plot_weights(f, figsize=(20, 4), return_fig=True)
        f_fig.tight_layout()
        rc_fig = viz_sequence.plot_weights(rc, figsize=(20, 4), return_fig=True)
        rc_fig.tight_layout()

        body.append(
            vdomh.tr(
                vdomh.td(str(j + 1)),
                vdomh.td(str(num_seqlets[i][j])),
                vdomh.td(util.figure_to_vdom_image(f_fig)),
                vdomh.td(util.figure_to_vdom_image(rc_fig))
            )
        )
    display(vdomh.table(colgroup, header, vdomh.tbody(*body)))
    plt.close("all")



    















    







Metacluster 1/2










    







#SeqletsForwardReverse
111154
2320
341










    







Metacluster 2/2










    







#SeqletsForwardReverse
1820
2562
3507
4172
5170
6101
7100
859
939
1036























Top TOMTOM matches for each motif
Here, the TF-MoDISco motifs are plotted as hCWMs, but the TOMTOM matches are shown as PWMs.













In [17]:



    


num_matches_to_keep = 10
num_matches_to_show = 5

header = vdomh.thead(
    vdomh.tr(
        vdomh.th("Motif ID", style={"text-align": "center"}),
        vdomh.th("q-val", style={"text-align": "center"}),
        vdomh.th("PWM", style={"text-align": "center"})
    )
)

for i in range(len(motif_pfms)):
    display(vdomh.h3("Metacluster %d/%d" % (i + 1, num_metaclusters)))
    
    # Compute TOMTOM matches for all motifs in the metacluster at once
    tomtom_matches = match_motifs_to_database(motif_pfms_short[i], top_k=num_matches_to_keep)
    
    for j in range(len(motif_pfms[i])):
        display(vdomh.h4("Motif %d/%d" % (j + 1, len(motif_pfms[i]))))
        viz_sequence.plot_weights(motif_hcwms[i][j])
    
        body = []
        for k, (match_name, match_pfm, match_qval) in enumerate(tomtom_matches[j]):
            fig = viz_sequence.plot_weights(util.pfm_to_pwm(match_pfm), return_fig=True)
            fig.tight_layout()
            if k < num_matches_to_show:
                body.append(
                    vdomh.tr(
                        vdomh.td(match_name),
                        vdomh.td(str(match_qval)),
                        vdomh.td(util.figure_to_vdom_image(fig))
                    )
                )
            else:
                body.append(
                    vdomh.tr(
                        vdomh.td(match_name),
                        vdomh.td(str(match_qval)),
                        vdomh.td("Not shown")
                    )
                )
        if not body:
            display(vdomh.p("No TOMTOM matches passing threshold"))
        else:
            display(vdomh.table(header, vdomh.tbody(*body)))
        plt.close("all")



    















    







Metacluster 1/2










    







Motif 1/3










    

















    







Motif IDq-valPWM
MA0139.1_CTCF5.84557e-17
CTCF_HUMAN.H11MO.0.A6.2359e-14
CTCFL_HUMAN.H11MO.0.A1.80634e-07
MA1102.2_CTCFL8.40079e-05
MA1568.1_TCF21(var.2)0.117237
MA1638.1_HAND20.139307Not shown
SNAI1_HUMAN.H11MO.0.C0.199133Not shown
ZIC3_HUMAN.H11MO.0.B0.20952600000000002Not shown
ZIC2_HUMAN.H11MO.0.D0.336227Not shown
MA0155.1_INSM10.38007399999999997Not shown










    







Motif 2/3










    

















    







Motif IDq-valPWM
MA0139.1_CTCF0.490624
CTCF_HUMAN.H11MO.0.A0.490624










    







Motif 3/3










    

















    







Motif IDq-valPWM
CPEB1_HUMAN.H11MO.0.D6.32445e-05
MA1125.1_ZNF3840.0296126
PRDM6_HUMAN.H11MO.0.C0.0296126
FOXL1_HUMAN.H11MO.0.D0.0296126
FOXG1_HUMAN.H11MO.0.D0.045944
ANDR_HUMAN.H11MO.0.A0.109287Not shown
FOXJ3_HUMAN.H11MO.0.A0.11081400000000001Not shown
MA0679.2_ONECUT10.11081400000000001Not shown
FUBP1_HUMAN.H11MO.0.D0.158685Not shown
ONEC2_HUMAN.H11MO.0.D0.158685Not shown










    







Metacluster 2/2










    







Motif 1/10










    

















    







Motif IDq-valPWM
CTCF_HUMAN.H11MO.0.A7.83135e-11
MA0139.1_CTCF9.07815e-11
CTCFL_HUMAN.H11MO.0.A1.0735699999999998e-09
MA1102.2_CTCFL9.2666e-06
ZIC3_HUMAN.H11MO.0.B0.06249299999999999
ZIC2_HUMAN.H11MO.0.D0.121233Not shown
MA1628.1_Zic1::Zic20.121233Not shown
PLAL1_HUMAN.H11MO.0.D0.121233Not shown
MA1629.1_Zic20.121233Not shown
KLF8_HUMAN.H11MO.0.C0.121233Not shown










    







Motif 2/10










    

















    







Motif IDq-valPWM
PO3F3_HUMAN.H11MO.0.D0.00227392
FOXD2_HUMAN.H11MO.0.D0.00417497
HXC10_HUMAN.H11MO.0.D0.00428995
LMX1A_HUMAN.H11MO.0.D0.00454879
MA0681.2_PHOX2B0.00740402
FOXG1_HUMAN.H11MO.0.D0.00792837Not shown
ARI3A_HUMAN.H11MO.0.D0.0098781Not shown
PO3F4_HUMAN.H11MO.0.D0.018273400000000002Not shown
MA0679.2_ONECUT10.0226644Not shown
CPEB1_HUMAN.H11MO.0.D0.0270004Not shown










    







Motif 3/10










    

















    







Motif IDq-valPWM
SP2_HUMAN.H11MO.0.A4.27271e-06
SP1_HUMAN.H11MO.0.A4.31371e-05
TBX15_HUMAN.H11MO.0.D7.00775e-05
SP3_HUMAN.H11MO.0.B7.00775e-05
KLF3_HUMAN.H11MO.0.B0.00033835800000000003
ZFX_HUMAN.H11MO.1.A0.00187597Not shown
PATZ1_HUMAN.H11MO.0.C0.00205376Not shown
MA1653.1_ZNF1480.00369656Not shown
KLF16_HUMAN.H11MO.0.D0.00369656Not shown
MA1513.1_KLF150.00476856Not shown










    







Motif 4/10










    

















    







Motif IDq-valPWM
SP2_HUMAN.H11MO.0.A5.9918e-08
SP3_HUMAN.H11MO.0.B5.56314e-07
SP1_HUMAN.H11MO.0.A5.50182e-06
MA1650.1_ZBTB140.000615282
KLF16_HUMAN.H11MO.0.D0.000615282
MA1513.1_KLF150.000615282Not shown
KLF3_HUMAN.H11MO.0.B0.000615282Not shown
SP1_HUMAN.H11MO.1.A0.000615282Not shown
PATZ1_HUMAN.H11MO.0.C0.00169829Not shown
USF2_HUMAN.H11MO.0.A0.00193856Not shown










    







Motif 5/10










    

















    







Motif IDq-valPWM
SP2_HUMAN.H11MO.0.A2.24896e-08
SP1_HUMAN.H11MO.0.A8.7211e-08
SP3_HUMAN.H11MO.0.B8.7211e-08
KLF16_HUMAN.H11MO.0.D2.3502199999999997e-07
TBX15_HUMAN.H11MO.0.D1.5602200000000002e-06
PATZ1_HUMAN.H11MO.0.C3.53602e-06Not shown
MAZ_HUMAN.H11MO.0.A3.53602e-06Not shown
WT1_HUMAN.H11MO.0.C3.53602e-06Not shown
ZN467_HUMAN.H11MO.0.C4.00358e-06Not shown
VEZF1_HUMAN.H11MO.0.C4.143229999999999e-05Not shown










    







Motif 6/10










    

















    







Motif IDq-valPWM
CTCF_HUMAN.H11MO.0.A0.0773322
MA0139.1_CTCF0.0773322
CTCFL_HUMAN.H11MO.0.A0.10286500000000001
MYOD1_HUMAN.H11MO.0.A0.10286500000000001
GABPA_HUMAN.H11MO.0.A0.22997399999999998
ZN341_HUMAN.H11MO.1.C0.334915Not shown
ETV4_HUMAN.H11MO.0.B0.334915Not shown
MA1628.1_Zic1::Zic20.334915Not shown
MA1508.1_IKZF10.334915Not shown
MA0062.3_GABPA0.37257199999999996Not shown










    







Motif 7/10










    

















    







Motif IDq-valPWM
SP2_HUMAN.H11MO.0.A0.000493407
SP3_HUMAN.H11MO.0.B0.00504961
SP1_HUMAN.H11MO.0.A0.00548569
ZFX_HUMAN.H11MO.1.A0.011178200000000001
USF2_HUMAN.H11MO.0.A0.0191305
PATZ1_HUMAN.H11MO.0.C0.026742200000000004Not shown
MA1650.1_ZBTB140.0522894Not shown
AP2D_HUMAN.H11MO.0.D0.0522894Not shown
SP4_HUMAN.H11MO.0.A0.0775836Not shown
RFX1_HUMAN.H11MO.0.B0.0775836Not shown










    







Motif 8/10










    

















    







Motif IDq-valPWM
SP2_HUMAN.H11MO.0.A4.4850200000000006e-05
SP3_HUMAN.H11MO.0.B6.14445e-05
TBX15_HUMAN.H11MO.0.D0.000187128
SP1_HUMAN.H11MO.0.A0.000187128
PATZ1_HUMAN.H11MO.0.C0.00022842200000000002
WT1_HUMAN.H11MO.0.C0.000294062Not shown
KLF3_HUMAN.H11MO.0.B0.000294062Not shown
ZN467_HUMAN.H11MO.0.C0.000716985Not shown
SP1_HUMAN.H11MO.1.A0.000758797Not shown
VEZF1_HUMAN.H11MO.0.C0.00123184Not shown










    







Motif 9/10










    

















    







Motif IDq-valPWM
SP1_HUMAN.H11MO.0.A0.0318789
CTCFL_HUMAN.H11MO.0.A0.0350778
TAL1_HUMAN.H11MO.0.A0.045011800000000005
PATZ1_HUMAN.H11MO.0.C0.130641
MA1653.1_ZNF1480.130641
MA0528.2_ZNF2630.130641Not shown
KLF16_HUMAN.H11MO.0.D0.130641Not shown
E2F6_HUMAN.H11MO.0.A0.130641Not shown
MA1102.2_CTCFL0.168958Not shown
CTCF_HUMAN.H11MO.0.A0.168958Not shown










    







Motif 10/10










    

















    







Motif IDq-valPWM
MA0753.2_ZNF7400.354917
KLF3_HUMAN.H11MO.0.B0.354917
MA0528.2_ZNF2630.354917
PATZ1_HUMAN.H11MO.0.C0.354917
MA0516.2_SP20.354917
MA0814.2_TFAP2C(var.2)0.354917Not shown
SP4_HUMAN.H11MO.1.A0.354917Not shown
KLF14_HUMAN.H11MO.0.D0.354917Not shown
ZN740_HUMAN.H11MO.0.D0.354917Not shown
PRDM4_HUMAN.H11MO.0.D0.354917Not shown























Sample of seqlets supporting each motif
Here, the motifs are presented as hCWMs, along with the actual importance scores of a random sample of seqlets that support the motif.













In [18]:



    


num_seqlets_to_show = 10

colgroup = vdomh.colgroup(
    vdomh.col(style={"width": "50%"}),
    vdomh.col(style={"width": "50%"})
)

header = vdomh.thead(
    vdomh.tr(
        vdomh.th("Motif hCWM", style={"text-align": "center"}),
        vdomh.th("Seqlets", style={"text-align": "center"})
    )
)

for i in range(len(motif_hcwms)):
    display(vdomh.h3("Metacluster %d/%d" % (i + 1, num_metaclusters)))
    
    for j in range(len(motif_hcwms[i])):
        display(vdomh.h4("Motif %d/%d" % (j + 1, len(motif_hcwms[i]))))
        
        motif_fig = viz_sequence.plot_weights(motif_hcwms[i][j], figsize=(20, 4), return_fig=True)
        motif_fig.tight_layout()
        
        seqlet_seqs, seqlet_hyps = motif_seqlets[i][j]
        
        sample_size = min(num_seqlets_to_show, len(seqlet_seqs))
        sample_inds = np.random.choice(len(seqlet_seqs), size=sample_size, replace=False)
        sample = []
        for k in sample_inds:
            fig = viz_sequence.plot_weights(seqlet_hyps[k] * seqlet_seqs[k], subticks_frequency=10, return_fig=True)
            fig.tight_layout()
            sample.append(util.figure_to_vdom_image(fig))
        body = vdomh.tbody(vdomh.tr(vdomh.td(util.figure_to_vdom_image(motif_fig)), vdomh.td(*sample)))
        display(vdomh.table(colgroup, header, body))
        plt.close("all")



    















    







Metacluster 1/2










    







Motif 1/3










    







Motif hCWMSeqlets










    







Motif 2/3










    







Motif hCWMSeqlets










    







Motif 3/3










    







Motif hCWMSeqlets










    







Metacluster 2/2










    







Motif 1/10










    







Motif hCWMSeqlets










    







Motif 2/10










    







Motif hCWMSeqlets










    







Motif 3/10










    







Motif hCWMSeqlets










    







Motif 4/10










    







Motif hCWMSeqlets










    







Motif 5/10










    







Motif hCWMSeqlets










    







Motif 6/10










    







Motif hCWMSeqlets










    







Motif 7/10










    







Motif hCWMSeqlets










    







Motif 8/10










    







Motif hCWMSeqlets










    







Motif 9/10










    







Motif hCWMSeqlets










    







Motif 10/10










    







Motif hCWMSeqlets