view rmsd_clustering.py @ 2:e0ecaf2d05fb draft default tip

"planemo upload for repository https://github.com/galaxycomputationalchemistry/galaxy-tools-compchem/ commit f1c3c88c7395f2e84cbc533199406aadb79c5c07"
author chemteam
date Fri, 13 Nov 2020 19:39:27 +0000
parents ad49025ba90d
children
line wrap: on
line source

import argparse
import json

import matplotlib.pyplot as plt

import numpy as np

from scipy.cluster.hierarchy import cophenet, dendrogram, linkage
from scipy.spatial.distance import pdist


def json_to_np(fname, start=None, end=None):
    """
    Load json file and convert to numpy array
    """
    with open(fname) as f:
        k = json.load(f)
    print(np.array(k)[:, :, start:end].shape)
    return np.array(k)[:, :, start:end]


def flatten_tensor(tensor, normalize=True):
    """
    Flatten tensor to a 2D matrix along the time axis
    """
    av = np.mean(tensor, axis=(0, 1)) if normalize else 1
    return np.mean(tensor/av, axis=2)


def get_cluster_linkage_array(mat, clustering_method='average'):
    Z = linkage(mat, clustering_method)
    c, coph_dists = cophenet(Z, pdist(mat))
    print('Cophenetic correlation coefficient: {}'.format(c))
    return Z


def plot_dist_mat(mat, output, cmap='plasma'):
    """
    Plot distance matrix as heatmap
    """
    fig, ax = plt.subplots(1)
    p = ax.pcolormesh(mat, cmap=cmap)
    plt.xlabel('Trajectory number')
    plt.ylabel('Trajectory number')
    plt.colorbar(p)
    plt.draw()
    plt.savefig(output, format='png')


def plot_dendrogram(Z, output):
    # figure width scales with number of leaves
    plt.figure(figsize=(0.25 * Z.shape[0], 10))
    plt.title('Hierarchical Clustering Dendrogram')
    plt.xlabel('Trajectory index')
    plt.ylabel('distance')
    dendrogram(
        Z,
        leaf_rotation=90.,  # rotates the x axis labels
        leaf_font_size=8.,  # font size for the x axis labels
    )
    plt.savefig(output, format='png')


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument('--json', help='JSON input file (for 3D tensor).')
    parser.add_argument('--mat', help='Input tabular file (for 2D matrix).')
    parser.add_argument('--outp-mat', help='Tabular output file.')
    parser.add_argument('--Z', required=True,
                        help='File for cluster linkage array.')
    parser.add_argument('--dendrogram',
                        help="Path to the output dendrogram file")
    parser.add_argument('--heatmap',
                        help="Path to the output distance matrix file")
    parser.add_argument('--clustering-method', default='average',
                        choices=['single', 'complete', 'average',
                                 'centroid', 'median', 'ward', 'weighted'],
                        help="Method to use for clustering.")
    parser.add_argument('--cmap', type=str, default='plasma',
                        help="Matplotlib colormap to use"
                             "for plotting distance matrix.")
    parser.add_argument('--start', type=int,
                        help="First trajectory frame to"
                             "calculate distance matrix")
    parser.add_argument('--end', type=int,
                        help="Last trajectory frame to"
                             "calculate distance matrix")
    parser.add_argument('--normalize', action="store_true",
                        help="Normalize the RMSD variation over"
                             "the trajectories before averaging.")
    args = parser.parse_args()

    print(args)
    if args.json:
        tensor = json_to_np(args.json, args.start, args.end)
        mat = flatten_tensor(tensor, args.normalize)
        np.savetxt(args.outp_mat, mat)
    elif args.mat:
        mat = np.loadtxt(args.mat)
    else:
        print("Either --json or --mat must be specified.")
        exit(1)

    Z = get_cluster_linkage_array(mat, args.clustering_method)
    np.savetxt(args.Z, Z)

    if args.heatmap:
        plot_dist_mat(mat, args.heatmap, args.cmap)

    if args.dendrogram:
        plot_dendrogram(Z, args.dendrogram)


if __name__ == "__main__":
    main()