Skip to content

Commit

Permalink
small fix for navis exercise 2
Browse files Browse the repository at this point in the history
  • Loading branch information
@schlegelp
schlegelp committed Sep 23, 2024
1 parent 064243f commit 26b345e
Showing 1 changed file with 39 additions and 25 deletions.
64 changes: 39 additions & 25 deletions exercises/navis/navis_exercise_2.ipynb
View file
Original file line number Diff line number Diff line change
Expand Up @@ -598,8 +598,8 @@
"outputs": [],
"source": [
"# NAVis pre-computes two kinds of segments:\n",
"print('Linear segmnents that maximmize segment lengths:', len(n.segments))\n",
"print('Linear segmnents between branch/leaf nodes:', len(n.small_segments))\n",
"print(\"Linear segmnents that maximmize segment lengths:\", len(n.segments))\n",
"print(\"Linear segmnents between branch/leaf nodes:\", len(n.small_segments))\n",
"\n",
"# For the tortuosity we need to use the latter!\n",
"\n",
Expand Down Expand Up @@ -628,9 +628,7 @@
"node_locs = dict(zip(n.nodes.node_id, n.nodes[[\"x\", \"y\", \"z\"]].values))\n",
"dist_eucl = []\n",
"for seg in n.small_segments:\n",
" dist_eucl.append(\n",
" np.linalg.norm(node_locs[seg[0]] - node_locs[seg[-1]])\n",
" )\n",
" dist_eucl.append(np.linalg.norm(node_locs[seg[0]] - node_locs[seg[-1]]))\n",
"dist_eucl = np.array(dist_eucl)\n",
"\n",
"# Inspect the results\n",
Expand Down Expand Up @@ -676,7 +674,7 @@
"dist_geo = []\n",
"for seg in n.small_segments:\n",
" dist_geo.append(\n",
" nx.shortest_path_length(n.graph, source=seg[0], target=seg[-1], weight='weight')\n",
" nx.shortest_path_length(n.graph, source=seg[0], target=seg[-1], weight=\"weight\")\n",
" )\n",
"dist_geo = np.array(dist_geo)\n",
"dist_geo[:10]"
Expand Down Expand Up @@ -791,7 +789,7 @@
"from scipy.cluster.hierarchy import linkage, cut_tree\n",
"\n",
"# Drop neurons that don't have a t-type assigned\n",
"nl_w_ttype = nl[nl.t_type != 'nan']\n",
"nl_w_ttype = nl[nl.t_type != \"nan\"]\n",
"\n",
"# First we will drop neurons that have not t-type assigned\n",
"ivscc_w_ttype = ivscc[nl_w_ttype.id]\n",
Expand All @@ -801,23 +799,27 @@
"ivscc_filled = ivscc_w_ttype.fillna(0)\n",
"\n",
"# Normalize the data to be between 0 and 1\n",
"ivscc_scaled = (ivscc_filled + ivscc_filled.min()) / (ivscc_filled.max() - ivscc_filled.min())\n",
"ivscc_scaled = (ivscc_filled - ivscc_filled.min(axis=1).values.reshape(-1, 1)) / (\n",
" ivscc_filled.max(axis=1) - ivscc_filled.min(axis=1)\n",
").values.reshape(-1, 1)\n",
"\n",
"ivscc_scaled.fillna(0, inplace=True)\n",
"\n",
"# Calculate the Euclidean distance\n",
"dist = pdist(ivscc_scaled.T)\n",
"\n",
"# Perform hierarchical clustering\n",
"Z = linkage(dist, method='ward')\n",
"Z = linkage(dist, method=\"ward\")\n",
"\n",
"# Cut the tree to get the clusters\n",
"labels = cut_tree(Z, n_clusters=40).flatten()\n",
"\n",
"# Combine cluster labels and t-types into a dataframe\n",
"clusters = pd.DataFrame()\n",
"\n",
"clusters['id'] = nl_w_ttype.id\n",
"clusters['t_type'] = nl_w_ttype.t_type\n",
"clusters['cluster'] = labels\n",
"clusters[\"id\"] = nl_w_ttype.id\n",
"clusters[\"t_type\"] = nl_w_ttype.t_type\n",
"clusters[\"cluster\"] = labels\n",
"clusters.head()"
]
},
Expand All @@ -834,9 +836,9 @@
"confusion = pd.crosstab(clusters.t_type, clusters.cluster)\n",
"\n",
"# Sort rows and columns by hierarchical clustering\n",
"row_linkage = linkage(pdist(confusion), method='ward')\n",
"row_linkage = linkage(pdist(confusion), method=\"ward\")\n",
"row_order = leaves_list(row_linkage)\n",
"col_linkage = linkage(pdist(confusion.T), method='ward')\n",
"col_linkage = linkage(pdist(confusion.T), method=\"ward\")\n",
"col_order = leaves_list(col_linkage)\n",
"\n",
"# Apply the sorting\n",
Expand All @@ -853,13 +855,20 @@
"outputs": [],
"source": [
"# For each t-type, generate a color based on how the first label\n",
"first_labels = np.unique(confusion.index.map(lambda x: x.split(' ')[0]))\n",
"first_labels = np.unique(confusion.index.map(lambda x: x.split(\" \")[0]))\n",
"\n",
"cmap = {}\n",
"for t, base_color in zip(first_labels, sns.color_palette(\"tab10\", len(first_labels))):\n",
" # All labels starting with `t`\n",
" t_labels = [ix for ix in confusion.index if ix.startswith(t)]\n",
" cmap.update(dict(zip(t_labels,sns.light_palette(base_color, n_colors=len(t_labels) * 2)[::-1])))"
" cmap.update(\n",
" dict(\n",
" zip(\n",
" t_labels,\n",
" sns.light_palette(base_color, n_colors=len(t_labels) * 2)[::-1],\n",
" )\n",
" )\n",
" )"
]
},
{
Expand All @@ -870,26 +879,31 @@
"outputs": [],
"source": [
"# Plot the confusion matrix as scatter heatmap\n",
"confusion_pivot = confusion.unstack().reset_index().rename(columns={0: 'count'})\n",
"confusion_pivot = confusion.unstack().reset_index().rename(columns={0: \"count\"})\n",
"\n",
"\n",
"# Turn the cluster into a string - otherwise seaborn will re-order it\n",
"confusion_pivot = confusion_pivot.astype({'cluster': \"string\", 't_type': \"string\"})\n",
"confusion_pivot = confusion_pivot.astype({\"cluster\": \"string\", \"t_type\": \"string\"})\n",
"\n",
"g = sns.relplot(\n",
" data=confusion_pivot,\n",
" y=\"t_type\", x=\"cluster\", size=\"count\",\n",
" palette=cmap, hue=\"t_type\",\n",
" y=\"t_type\",\n",
" x=\"cluster\",\n",
" size=\"count\",\n",
" palette=cmap,\n",
" hue=\"t_type\",\n",
" edgecolor=\".7\",\n",
" col_order=confusion.index.values, row_order=confusion.columns.values,\n",
" height=10, sizes=(0, 400),\n",
" legend=False\n",
" col_order=confusion.index.values,\n",
" row_order=confusion.columns.values,\n",
" height=10,\n",
" sizes=(0, 400),\n",
" legend=False,\n",
")\n",
"\n",
"ax = g.axes[0, 0]\n",
"\n",
"ax.set_ylabel('Transcriptomic type')\n",
"ax.set_xlabel('Morphological cluster')"
"ax.set_ylabel(\"Transcriptomic type\")\n",
"ax.set_xlabel(\"Morphological cluster\")"
]
},
{
Expand Down

0 comments on commit 26b345e

Please sign in to comment.