minor changes to notebooks

Author: vivekbhr

docs/content/tutorials/GLM_PCA_analysis.ipynb renamed to docs/content/tutorials/snmCATseq_glmPCA_analysis.ipynb

@@ -5,10 +5,14 @@
    "id": "ee0fddbe-1c52-44c1-b2bc-b3f4165209cf",
    "metadata": {},
    "source": [
-    "# Tutorial for snmC2Tseq\n",
-    "Using the single-cell methylation data from snmC2Tseq [1], we re-processed the dataset to extract smaller bins (10kb), yielding non-Gaussian data. The data is losely following a Beta distribution, and we thus employ GLM-PCA with Beta distribution to find a lower-dimension representation.\n",
-    "<br/>\n",
-    "[1]: https://www.sciencedirect.com/science/article/pii/S2666979X22000271?via%3Dihub"
+    "# GLM-PCA analysis of single-cell methylome data\n",
+    "\n",
+    "\n",
+    "\n",
+    "Starting from the publicly available single-cell methylation data from [snmCATseq](https://www.sciencedirect.com/science/article/pii/S2666979X22000271?via%3Dihub) , we re-processed a subset of the data to extract methlated/unmethylated reads in small genomics bins (10kb), yielding non-Gaussian data. \n",
+    "\n",
+    "\n",
+    "The data is losely following a Beta distribution, and we thus employ GLM-PCA with Beta distribution to find a lower-dimension representation.\n"
    ]
   },
   {
@@ -40,7 +44,11 @@
    "metadata": {},
    "source": [
     "## Read data\n",
-    "Data needs to be processed as demonstrated in \"snmC2Tseq_prepcessing.ipynb\". We here load it."
+    "\n",
+    "\n",
+    "Starting from the files provided in our [figshare repository](), the data needs to be first processed as demonstrated in \"snmCATseq_prepcessing.ipynb\". We here load the output.\n",
+    "\n",
+    "In order to confirm whether glmPCA results help in identifying clusters of cells, we shall also load the celltype metadata provided in the Supplementary table 5 of the original manuscript.\n"
    ]
   },
   {
@@ -83,9 +91,13 @@
    "metadata": {},
    "source": [
     "## Sincei: GLM-PCA\n",
-    "Two Beta families have been designed: Beta and SigmoidBeta. SigmoidBeta employs a logit-transformed saturated parameters, which removes all optimisation constraints and is thus much more stable.\n",
-    "<br/>\n",
-    "Several hyper-parameters are hard-coded, but do not vastly influence the analysis. The only hyper-parameter which can change results fondamentally is the learning rate. 2.5 is a good choice for sigmoid_beta, but this needs to be changed for other distributions.2**8"
+    "\n",
+    "To aid in this analysis, we provide two Beta families within the glmPCA framework: **Beta** and **SigmoidBeta**. \n",
+    "\n",
+    "SigmoidBeta employs a logit-transformed saturated parameters, which removes all optimisation constraints and is therefore much more stable.\n",
+    "\n",
+    "\n",
+    "Several hyper-parameters are hard-coded, but do not vastly influence the analysis. The only hyper-parameter which can change results fondamentally is the learning rate. 2.5 is a good choice for sigmoid_beta, but this needs to be changed for other distributions."
    ]
   },
   {
@@ -175,19 +187,19 @@
     "_umap_params = {'n_neighbors':15, 'min_dist': 0.3, 'n_epochs': 1000}\n",
     "\n",
     "metric = 'cosine'\n",
+    "\n",
     "_umap_clf = umap.UMAP(metric=metric, verbose=True, **_umap_params)\n",
     "umap_embeddings = pd.DataFrame(\n",
     "    _umap_clf.fit_transform(X_project.detach().numpy()), \n",
     "    columns=['UMAP 1', 'UMAP 2']\n",
     ").reset_index()\n",
-    "umap_embeddings['label'] = cell_labels['snmC2T-seq Baseline Cluster'].values\n",
+    "umap_embeddings['label'] = cell_labels['snmCAT-seq Baseline Cluster'].values\n",
     "\n",
     "g = sns.relplot(data=umap_embeddings, x='UMAP 1', y='UMAP 2',hue='label')\n",
     "figure_name = 'UMAP_glm_pca_%s_metric_%s_%s%s'%(\n",
     "    n_pc, \n",
     "    metric,\n",
-    "    family,\n",
-    "    '_' + method if 'beta' in family else ''\n",
+    "    family\n",
     ")"
    ]
   },
@@ -266,9 +278,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "sincei",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
-   "name": "sincei"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {
@@ -280,7 +292,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.16"
+   "version": "3.8.16"
   }
  },
  "nbformat": 4,

docs/content/tutorials/snmC2Tseq_preprocessing.ipynb renamed to docs/content/tutorials/snmCATseq_preprocessing.ipynb

@@ -5,8 +5,12 @@
    "id": "f1dc587b-e8b8-4beb-ab8b-6010e774d52a",
    "metadata": {},
    "source": [
-    "# Pre-process snmC2Tseq data for later use in Percolate\n",
-    "<b>NOTES</b>:\n",
+    "# Pre-process snmCATseq data for later use in sincei\n",
+    "\n",
+    "This notebook uses a subset of data from the snmCAT-seq protocol presented in [Luo et. al (2022)](https://www.sciencedirect.com/science/article/pii/S2666979X22000271)\n",
+    "\n",
+    "\n",
+    "**NOTES**:\n",
     "- unmeth is not a proper column naming, it actually corresponds to the total number of reads. It can be seen as the ratio of meth/unmeth maxes out at 0.5."
    ]
   },
@@ -1352,9 +1356,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "sincei",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
-   "name": "sincei"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {
@@ -1366,7 +1370,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.16"
+   "version": "3.8.16"
   }
  },
  "nbformat": 4,