Merge branch 'master' of https://github.com/dpeerlab/Palantir

# Conflicts:
#	src/palantir/io.py

src/palantir/core.py

@@ -93,7 +93,9 @@ def run_palantir(
 
     # pseudotime and weighting matrix
     print("Determining pseudotime...")
-    pseudotime, W = _compute_pseudotime(data, start_cell, knn, waypoints, n_jobs, max_iterations)
+    pseudotime, W = _compute_pseudotime(
+        data, start_cell, knn, waypoints, n_jobs, max_iterations
+    )
 
     # Entropy and branch probabilities
     print("Entropy and branch probabilities...")
@@ -274,7 +276,9 @@ def identify_terminal_states(
     waypoints = pd.Index([start_cell]).append(waypoints)
 
     # Distance to start cell as pseudo pseudotime
-    pseudotime, _ = _compute_pseudotime(data, start_cell, knn, waypoints, n_jobs, max_iterations)
+    pseudotime, _ = _compute_pseudotime(
+        data, start_cell, knn, waypoints, n_jobs, max_iterations
+    )
 
     # Markov chain
     wp_data = data.loc[waypoints, :]

src/palantir/io.py

@@ -0,0 +1,124 @@
+import numpy as np
+import pandas as pd
+import os.path
+import fcsparser
+import scanpy as sc
+from scipy.io import mmread
+
+
+def _clean_up(df):
+    df = df.loc[df.index[df.sum(axis=1) > 0], :]
+    df = df.loc[:, df.columns[df.sum() > 0]]
+    return df
+
+
+def from_csv(counts_csv_file, delimiter=","):
+    # Read in csv file
+    df = pd.read_csv(counts_csv_file, sep=delimiter, index_col=0)
+    clean_df = _clean_up(df)
+    return clean_df
+
+
+def from_mtx(mtx_file, gene_name_file):
+
+    # Read in mtx file
+    count_matrix = mmread(mtx_file)
+
+    gene_names = np.loadtxt(gene_name_file, dtype=np.dtype("S"))
+    gene_names = np.array([gene.decode("utf-8") for gene in gene_names])
+
+    # remove todense
+    df = pd.DataFrame(count_matrix.todense(), columns=gene_names)
+
+    return _clean_up(df)
+
+
+def from_10x(data_dir, use_ensemble_id=True):
+    # loads 10x sparse format data
+    # data_dir is dir that contains matrix.mtx, genes.tsv and barcodes.tsv
+    # return_sparse=True -- returns data matrix in sparse format (default = False)
+
+    if data_dir is None:
+        data_dir = "./"
+    elif data_dir[len(data_dir) - 1] != "/":
+        data_dir = data_dir + "/"
+
+    filename_dataMatrix = os.path.expanduser(data_dir + "matrix.mtx")
+    filename_genes = os.path.expanduser(data_dir + "genes.tsv")
+    filename_cells = os.path.expanduser(data_dir + "barcodes.tsv")
+
+    # Read in gene expression matrix (sparse matrix)
+    # Rows = genes, columns = cells
+    dataMatrix = mmread(filename_dataMatrix)
+
+    # Read in row names (gene names / IDs)
+    gene_names = np.loadtxt(filename_genes, delimiter="\t", dtype=bytes).astype(str)
+    if use_ensemble_id:
+        gene_names = [gene[0] for gene in gene_names]
+    else:
+        gene_names = [gene[1] for gene in gene_names]
+    cell_names = np.loadtxt(filename_cells, delimiter="\t", dtype=bytes).astype(str)
+
+    dataMatrix = pd.DataFrame(
+        dataMatrix.todense(), columns=cell_names, index=gene_names
+    )
+
+    # combine duplicate genes
+    if not use_ensemble_id:
+        dataMatrix = dataMatrix.groupby(dataMatrix.index).sum()
+    dataMatrix = dataMatrix.transpose()
+
+    return _clean_up(dataMatrix)
+
+
+def from_10x_HDF5(filename, genome=None):
+
+    ad = sc.read_10x_h5(filename, genome, True)
+
+    dataMatrix = pd.DataFrame(ad.X.todense(), columns=ad.var_names, index=ad.obs_names)
+
+    return _clean_up(dataMatrix)
+
+
+def from_fcs(
+    cls,
+    fcs_file,
+    cofactor=5,
+    metadata_channels=[
+        "Time",
+        "Event_length",
+        "DNA1",
+        "DNA2",
+        "Cisplatin",
+        "beadDist",
+        "bead1",
+    ],
+):
+
+    # Parse the fcs file
+    text, data = fcsparser.parse(fcs_file)
+    data = data.astype(np.float64)
+
+    # Extract the S and N features (Indexing assumed to start from 1)
+    # Assumes channel names are in S
+    no_channels = text["$PAR"]
+    channel_names = [""] * no_channels
+    for i in range(1, no_channels + 1):
+        # S name
+        try:
+            channel_names[i - 1] = text["$P%dS" % i]
+        except KeyError:
+            channel_names[i - 1] = text["$P%dN" % i]
+    data.columns = channel_names
+
+    # Metadata and data
+    metadata_channels = data.columns.intersection(metadata_channels)
+    data_channels = data.columns.difference(metadata_channels)
+    # metadata = data[metadata_channels]
+    data = data[data_channels]
+
+    # Transform if necessary
+    if cofactor is not None or cofactor > 0:
+        data = np.arcsinh(np.divide(data, cofactor))
+
+    return data

src/palantir/plot.py

@@ -363,7 +363,7 @@ def plot_palantir_results(pr_res, tsne):
 
     for i, branch in enumerate(pr_res.branch_probs.columns):
         row = int(np.floor(i / n_cols))
-        ax = plt.subplot(gs[row + 2, np.remainder(i,n_cols)])
+        ax = plt.subplot(gs[row + 2, np.remainder(i, n_cols)])
         c = pr_res.branch_probs.loc[tsne.index, branch]
         ax.scatter(
             tsne.loc[:, "x"], tsne.loc[:, "y"], s=3, cmap=matplotlib.cm.plasma, c=c