{ "cells": [ { "cell_type": "markdown", "id": "34ddfdec", "metadata": {}, "source": [ "# Getting started: validating a drug target with TwinCell\n", "\n", "### Question: *Is Deucravacitinib an interesting drug for psoriasis?*\n", "\n", "**Deucravacitinib** is an oral, **selective TYK2 inhibitor** approved for moderate-to-severe\n", "plaque psoriasis. TYK2 relays **IL-23, IL-12 and type-I interferon** signalling through the\n", "JAK–STAT pathway — a central axis of psoriasis — so it is a strong candidate to investigate.\n", "\n", "Rather than running a wet-lab experiment, we use **TwinCell** to ask a causal question\n", "directly from patient single-cell data:\n", "\n", "> Given the genes that change in psoriasis, does inhibiting **Deucravacitinib's target (TYK2)**\n", "> causally explain that disease signature?\n", "\n", "A high **target-validation score** means the target sits upstream of many of the\n", "disease-associated genes in the interactome — i.e. modulating it is mechanistically\n", "plausible for reversing the disease state.\n", "\n", "**Workflow at a glance**\n", "\n", "1. Load psoriatic vs. healthy skin single-cell data.\n", "2. Aggregate cells into **pseudo-bulk** profiles per sample.\n", "3. Compute the **psoriasis signature** (differentially expressed genes) with PyDESeq2.\n", "4. Open a **TwinCell** session and **validate** TYK2 against that signature.\n", "5. Inspect the **causal graph and paths** that link the disease genes to the targets.\n", "\n", "> You need a `DEEPLIFE_API_KEY` (from the TwinCell Console). Put it in a local `.env`\n", "> file or your environment before running the notebook." ] }, { "cell_type": "markdown", "id": "94ef4e1c", "metadata": {}, "source": [ "## 1. Setup\n", "\n", "Import the TwinCell SDK helpers and load your API key." ] }, { "cell_type": "code", "execution_count": 1, "id": "814ce6c9", "metadata": {}, "outputs": [], "source": [ "import os\n", "\n", "import matplotlib.pyplot as plt\n", "import pandas as pd\n", "import seaborn as sns\n", "from dotenv import find_dotenv, load_dotenv\n", "from IPython.display import display\n", "\n", "from deeplife.twincell import (\n", " TwinCell,\n", " pseudobulk,\n", " pydeseq2,\n", " read_h5ad,\n", ")\n", "\n", "# Load DEEPLIFE_API_KEY from a local .env file (walks up from the notebook's cwd).\n", "load_dotenv(find_dotenv(usecwd=True), override=True)\n", "DEEPLIFE_API_KEY = os.environ.get(\"DEEPLIFE_API_KEY\", \"dl-xxx\")" ] }, { "cell_type": "markdown", "id": "0c9155fb", "metadata": {}, "source": [ "## 2. The drug and its targets\n", "\n", "TwinCell validates **protein targets**, not drug names. Deucravacitinib selectively inhibits\n", "**TYK2**, so that is the target we will test against the psoriasis signature." ] }, { "cell_type": "code", "execution_count": 2, "id": "60b7581c", "metadata": {}, "outputs": [], "source": [ "# Deucravacitinib's primary protein target (HGNC symbol).\n", "deucravacitinib_targets = [\"TYK2\"]" ] }, { "cell_type": "markdown", "id": "9c317e43", "metadata": {}, "source": [ "## 3. Load the single-cell data\n", "\n", "We use a public psoriasis skin atlas ([GSE162183](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE162183)),\n", "which contains both **psoriatic** and **healthy (normal)** skin biopsies annotated by cell\n", "type. `read_h5ad` accepts local paths as well as `s3://` / `https://` URIs.\n", "\n", "TwinCell expects **HGNC gene symbols** as the variable index, so we set them up here." ] }, { "cell_type": "code", "execution_count": 3, "id": "9c18c275", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "View of AnnData object with n_obs × n_vars = 24126 × 28080\n", " obs: 'donor_id', 'Species', 'CellTypes', 'sample_id', 'protocol_url', 'institute', 'tissue_ontology_term_id', 'tissue_type', 'assay_ontology_term_id', 'cell_type_ontology_term_id', 'self_reported_ethnicity_ontology_term_id', 'development_stage_ontology_term_id', 'sex_ontology_term_id', 'suspension_type', 'disease_ontology_term_id', 'library_id', 'library_preparation_batch', 'library_sequencing_run', 'alignment_software', 'manner_of_death', 'sample_source', 'sample_collection_method', 'sample_preservation_method', 'sequenced_fragment', 'reference_genome', 'cell_enrichment', 'gene_annotation_version', 'is_primary_data', 'sampled_site_condition', 'cell_type', 'assay', 'disease', 'sex', 'tissue', 'self_reported_ethnicity', 'development_stage', 'observation_joinid'\n", " var: 'GeneSym', 'feature_is_filtered', 'feature_reference', 'feature_biotype', 'feature_length', 'feature_type'\n", " uns: 'batch_condition', 'citation', 'is_pre_analysis', 'organism', 'organism_ontology_term_id', 'schema_reference', 'schema_version', 'study_pi', 'title'\n", " obsm: 'X_umap'" ] }, "execution_count": 3, "metadata": {}, "output_type": "execute_result" } ], "source": [ "adata = read_h5ad(\n", " \"s3://dl-dev-global-personal/jbmorlot/cellxgenes/raw/adata_psoriasis_cxg_GSE162183.h5ad\"\n", ")\n", "\n", "# Use HGNC gene symbols as the variable index, and drop duplicate symbols.\n", "adata.var.rename(columns={\"feature_name\": \"gene_name\"}, inplace=True)\n", "adata.var.set_index(\"gene_name\", inplace=True)\n", "adata = adata[:, ~adata.var.index.duplicated()]\n", "adata" ] }, { "cell_type": "markdown", "id": "d3f18a86", "metadata": {}, "source": [ "Three `obs` columns drive the analysis:\n", "\n", "- **`disease`** — the condition (`psoriasis` vs `normal`); this is the perturbation.\n", "- **`cell_type`** — used to analyse a single, homogeneous population.\n", "- **`sample_id`** — the biological replicate id used to build pseudo-bulk profiles." ] }, { "cell_type": "code", "execution_count": 4, "id": "138f9477", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
diseasecell_typecount
38normalLangerhans cell21
37psoriasisLangerhans cell22
22psoriasisbasal cell of epidermis207
33normalbasal cell of epidermis50
12psoriasiscytotoxic T cell697
13normalcytotoxic T cell644
21psoriasisdendritic cell, human237
19normaldendritic cell, human238
3psoriasisendothelial cell1923
7normalendothelial cell1315
26psoriasisendothelial cell of lymphatic vessel163
23normalendothelial cell of lymphatic vessel197
41normalerythrocyte7
39psoriasiserythrocyte19
16psoriasisfibroblast442
18normalfibroblast400
9normalfibroblast of papillary layer of dermis887
8psoriasisfibroblast of papillary layer of dermis917
15normalgranular cell of epidermis462
11psoriasisgranular cell of epidermis706
17psoriasishair follicular keratinocyte409
31normalhair follicular keratinocyte117
27psoriasishelper T cell163
25normalhelper T cell181
30psoriasismacrophage117
20normalmacrophage238
28normalmelanocyte of skin155
34psoriasismelanocyte of skin49
36normalmonocyte26
24psoriasismonocyte187
40normalnatural killer cell14
35psoriasisnatural killer cell49
5normalpericyte1560
2psoriasispericyte2140
29psoriasisregulatory T cell119
32normalregulatory T cell63
14psoriasisskin fibroblast495
10normalskin fibroblast858
6psoriasisspinous cell of epidermis1341
1normalspinous cell of epidermis2343
4normalsuprabasal keratinocyte1593
0psoriasissuprabasal keratinocyte2355
\n", "
" ], "text/plain": [ " disease cell_type count\n", "38 normal Langerhans cell 21\n", "37 psoriasis Langerhans cell 22\n", "22 psoriasis basal cell of epidermis 207\n", "33 normal basal cell of epidermis 50\n", "12 psoriasis cytotoxic T cell 697\n", "13 normal cytotoxic T cell 644\n", "21 psoriasis dendritic cell, human 237\n", "19 normal dendritic cell, human 238\n", "3 psoriasis endothelial cell 1923\n", "7 normal endothelial cell 1315\n", "26 psoriasis endothelial cell of lymphatic vessel 163\n", "23 normal endothelial cell of lymphatic vessel 197\n", "41 normal erythrocyte 7\n", "39 psoriasis erythrocyte 19\n", "16 psoriasis fibroblast 442\n", "18 normal fibroblast 400\n", "9 normal fibroblast of papillary layer of dermis 887\n", "8 psoriasis fibroblast of papillary layer of dermis 917\n", "15 normal granular cell of epidermis 462\n", "11 psoriasis granular cell of epidermis 706\n", "17 psoriasis hair follicular keratinocyte 409\n", "31 normal hair follicular keratinocyte 117\n", "27 psoriasis helper T cell 163\n", "25 normal helper T cell 181\n", "30 psoriasis macrophage 117\n", "20 normal macrophage 238\n", "28 normal melanocyte of skin 155\n", "34 psoriasis melanocyte of skin 49\n", "36 normal monocyte 26\n", "24 psoriasis monocyte 187\n", "40 normal natural killer cell 14\n", "35 psoriasis natural killer cell 49\n", "5 normal pericyte 1560\n", "2 psoriasis pericyte 2140\n", "29 psoriasis regulatory T cell 119\n", "32 normal regulatory T cell 63\n", "14 psoriasis skin fibroblast 495\n", "10 normal skin fibroblast 858\n", "6 psoriasis spinous cell of epidermis 1341\n", "1 normal spinous cell of epidermis 2343\n", "4 normal suprabasal keratinocyte 1593\n", "0 psoriasis suprabasal keratinocyte 2355" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "perturbation_col = \"disease\" # \"psoriasis\" vs \"normal\"\n", "cell_type_col = \"cell_type\"\n", "batch_id_col = \"sample_id\" # pseudo-bulk replicate id\n", "\n", "adata.obs[[perturbation_col, cell_type_col]].value_counts().reset_index().sort_values(by=\"cell_type\")" ] }, { "cell_type": "markdown", "id": "9e095e36", "metadata": {}, "source": [ "## 4. Pseudo-bulk aggregation\n", "\n", "Single cells are noisy and statistically dependent within a sample. `pseudobulk` sums counts\n", "per **`sample_id` × `cell_type` × `disease`** group, producing a handful of robust replicates\n", "per condition — the right input for differential expression. Groups with fewer than\n", "`n_min_replicates` cells are dropped." ] }, { "cell_type": "code", "execution_count": 5, "id": "71c07a7c", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "Building pseudo-bulk: 0%| | 0/125 [00:00" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "de_results = pydeseq2(\n", " adata=pdata_control.concatenate(pdata_pert),\n", " design_factor=perturbation_col,\n", " control_group=control,\n", " log2fc_sig=1.0,\n", " mlog10pvalue_sig=1.3,\n", ")\n", "\n", "significant_degs = de_results[de_results[\"significant\"]].index.tolist()\n", "print(f\"{len(significant_degs)} significant DEGs (psoriasis vs normal)\")\n", "\n", "sns.scatterplot(\n", " data=de_results, x=\"log2FoldChange\", y=\"mlog10pvalue_adj\", hue=\"significant\"\n", ")\n", "plt.xlabel(\"log2 fold change (psoriasis vs normal)\")\n", "plt.ylabel(\"-log10 adjusted p-value\")\n", "plt.title(f\"Psoriasis signature in {cell_type}\")\n", "plt.show()" ] }, { "cell_type": "markdown", "id": "f2d6600e", "metadata": {}, "source": [ "## 7. Open a TwinCell session\n", "\n", "A `TwinCell` session bundles the control arm, the perturbed arm, and the disease signature.\n", "On construction the SDK **validates** the inputs (shared genes, counts in `.X`, HGNC-style\n", "DEG symbols) and checks **API connectivity** — look for a `TwinCell: OK ...` line in the\n", "output." ] }, { "cell_type": "code", "execution_count": 9, "id": "56b1b595", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "pdata_control and pdata_pert share identical var_names order.\n", "DEG list: 131 submitted, 129 HGNC-style (usable by TwinCell), 2 not HGNC-style (ignored at inference; e.g. Ensembl ids). \n", "TwinCell: OK - input data ready for prediction (3 control and 3 perturbed observations, 131 DEGs submitted, 129 HGNC-style usable); API reachable at 'https://open-deeplife-api.deeplife.co'.\n" ] } ], "source": [ "tc = TwinCell(\n", " pdata_control=pdata_control,\n", " pdata_pert=pdata_pert,\n", " degs=significant_degs,\n", " model=\"v1.0\",\n", " api_key=DEEPLIFE_API_KEY,\n", ")" ] }, { "cell_type": "markdown", "id": "7ba1f1b5", "metadata": {}, "source": [ "## 8. Validate Deucravacitinib's target\n", "\n", "This is the heart of the analysis. For each target we call `target_validation`, which submits\n", "the two arms + signature and runs the causal model server-side. `get_target_score` then\n", "returns:\n", "\n", "- **`score`** — how strongly the target explains the disease signature (higher = stronger).\n", "- **`percentage_degs_significant`** — fraction of the disease DEGs with causal support from\n", " this target.\n", "\n", "We run target validation for **TYK2** (the loop also lets you add more targets to compare in a\n", "single table)." ] }, { "cell_type": "code", "execution_count": null, "id": "5e024340", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "status=pending\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "TwinCellSession prediction_id: f8e709d4-72b8-4c67-81e6-4d0412744625\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "status=processing · step=inference_started · 18%\n", "status=completed · step=completed · 100%\n" ] }, { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
idscorerankrank_percentagepvaluez_scoreprotein_expressedis_degnode_typen_degs_significantpercentage_degs_significant
0TYK2|PROTEIN22.708529575.04.264313NoneNoneTrueFalsePROTEIN21.680672
\n", "
" ], "text/plain": [ " id score rank rank_percentage pvalue z_score \\\n", "0 TYK2|PROTEIN 22.708529 575.0 4.264313 None None \n", "\n", " protein_expressed is_deg node_type n_degs_significant \\\n", "0 True False PROTEIN 2 \n", "\n", " percentage_degs_significant \n", "0 1.680672 " ] }, "execution_count": 10, "metadata": {}, "output_type": "execute_result" } ], "source": [ "prediction_ids = {}\n", "score_rows = []\n", "\n", "# Run target validation for each Deucravacitinib target - ETA 40s\n", "for target in deucravacitinib_targets:\n", " prediction_ids[target] = tc.target_validation(\n", " target=target,\n", " label=f\"Deucravacitinib target {target} vs psoriasis signature\",\n", " )\n", " score_rows.append(tc.get_target_score(prediction_id=prediction_ids[target]))\n", "\n", "target_scores = pd.concat(score_rows, ignore_index=True)\n", "target_scores" ] }, { "cell_type": "markdown", "id": "9da9b831", "metadata": {}, "source": [ "## 9. How do the targets connect to the disease genes?\n", "\n", "A score is more convincing when you can see the mechanism. `plot_causal_graph` draws the\n", "interactome subgraph linking the top disease DEGs to the validated target." ] }, { "cell_type": "code", "execution_count": 11, "id": "cb08d7e9", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Best target: TYK2|PROTEIN\n" ] }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Pick the highest-scoring validated target.\n", "best_target = target_scores.sort_values(\"score\", ascending=False)[\"id\"].iloc[0]\n", "best_prediction_id = prediction_ids[best_target.split(\"|\")[0]]\n", "print(f\"Best target: {best_target}\")\n", "\n", "n_sig_degs = target_scores[target_scores['id']==best_target].n_degs_significant\n", "\n", "tc.plot_causal_graph(prediction_id=best_prediction_id, top_n_degs=n_sig_degs)" ] }, { "cell_type": "markdown", "id": "f2b4f235", "metadata": {}, "source": [ "## 10. Inspect the top causal paths\n", "\n", "`get_causal_paths` returns the ranked routes from each impacted disease gene to the target.\n", "For a TYK2 target you should see paths funnelling through **STAT** transcription factors and\n", "other IL-23 / type-I-interferon JAK–STAT components — exactly the biology Deucravacitinib is\n", "designed to dampen." ] }, { "cell_type": "code", "execution_count": 12, "id": "b2910d47", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
degtargetpathscorecoveragepath_fractionlengthnodespath_probabilityscore_deg_given_target
0IL15TYK2|PROTEINIL15 → IRF1 → NPM1 → TYK20.7685841.00.9494644[IL15|RNA, IRF1|PROTEIN, NPM1|PROTEIN, TYK2|PR...0.0075450.809492
1IFITM3TYK2|PROTEINIFITM3 → STAT2 → TYK20.7637761.00.8592083[IFITM3|RNA, STAT2|PROTEIN, TYK2|PROTEIN]0.0086160.888930
2ZC3H12ATYK2|PROTEINZC3H12A → RELA → HRH1 → GRK2 → GNB2 → PTAFR → ...0.5261581.00.9579417[ZC3H12A|RNA, RELA|PROTEIN, HRH1|PROTEIN, GRK2...0.0022270.549259
3ASGR2TYK2|PROTEINASGR2 → IRF1 → NPM1 → TYK20.4557861.00.7228574[ASGR2|RNA, IRF1|PROTEIN, NPM1|PROTEIN, TYK2|P...0.0025660.630534
4RASGEF1BTYK2|PROTEINRASGEF1B → STAT1 → TYK20.4210861.00.9508783[RASGEF1B|RNA, STAT1|PROTEIN, TYK2|PROTEIN]0.0017810.442840
5INSIG1TYK2|PROTEININSIG1 → CCNT2 → HEXIM1 → NPM1 → TYK20.3526541.00.8538025[INSIG1|RNA, CCNT2|PROTEIN, HEXIM1|PROTEIN, NP...0.0014080.413040
6KIF2ATYK2|PROTEINKIF2A → TCF4 → CSNK2A1 → THAP1 → STRBP → EIF2A...0.3368331.00.6955277[KIF2A|RNA, TCF4|PROTEIN, CSNK2A1|PROTEIN, THA...0.0006120.484285
7LAP3TYK2|PROTEINLAP3 → STAT1 → TYK20.3132561.00.5006293[LAP3|RNA, STAT1|PROTEIN, TYK2|PROTEIN]0.0019830.625725
8LAP3TYK2|PROTEINLAP3 → STAT2 → TYK20.2962861.00.4735083[LAP3|RNA, STAT2|PROTEIN, TYK2|PROTEIN]0.0018760.625725
9IL1R1TYK2|PROTEINIL1R1 → TCF4 → CSNK2A1 → THAP1 → STRBP → EIF2A...0.2729551.00.5103127[IL1R1|RNA, TCF4|PROTEIN, CSNK2A1|PROTEIN, THA...0.0005800.534878
10HES1TYK2|PROTEINHES1 → EGR1 → RELA → HRH1 → GRK2 → GNB2 → PTAFR0.2176611.00.5198037[HES1|RNA, EGR1|PROTEIN, RELA|PROTEIN, HRH1|PR...0.0003270.418738
11ASGR2TYK2|PROTEINASGR2 → EGR1 → RELA → HRH1 → GRK2 → GNB2 → PTAFR0.1681131.00.2666207[ASGR2|RNA, EGR1|PROTEIN, RELA|PROTEIN, HRH1|P...0.0009470.630534
12HES1TYK2|PROTEINHES1 → TCF4 → CSNK2A1 → THAP1 → STRBP → EIF2AK...0.1439881.00.3438627[HES1|RNA, TCF4|PROTEIN, CSNK2A1|PROTEIN, THAP...0.0002160.418738
13IFITM3TYK2|PROTEINIFITM3 → STAT1 → TYK20.1190061.00.1338763[IFITM3|RNA, STAT1|PROTEIN, TYK2|PROTEIN]0.0013430.888930
14IL1R1TYK2|PROTEINIL1R1 → STAT1 → TYK20.1139681.00.2130733[IL1R1|RNA, STAT1|PROTEIN, TYK2|PROTEIN]0.0002420.534878
\n", "
" ], "text/plain": [ " deg target path \\\n", "0 IL15 TYK2|PROTEIN IL15 → IRF1 → NPM1 → TYK2 \n", "1 IFITM3 TYK2|PROTEIN IFITM3 → STAT2 → TYK2 \n", "2 ZC3H12A TYK2|PROTEIN ZC3H12A → RELA → HRH1 → GRK2 → GNB2 → PTAFR → ... \n", "3 ASGR2 TYK2|PROTEIN ASGR2 → IRF1 → NPM1 → TYK2 \n", "4 RASGEF1B TYK2|PROTEIN RASGEF1B → STAT1 → TYK2 \n", "5 INSIG1 TYK2|PROTEIN INSIG1 → CCNT2 → HEXIM1 → NPM1 → TYK2 \n", "6 KIF2A TYK2|PROTEIN KIF2A → TCF4 → CSNK2A1 → THAP1 → STRBP → EIF2A... \n", "7 LAP3 TYK2|PROTEIN LAP3 → STAT1 → TYK2 \n", "8 LAP3 TYK2|PROTEIN LAP3 → STAT2 → TYK2 \n", "9 IL1R1 TYK2|PROTEIN IL1R1 → TCF4 → CSNK2A1 → THAP1 → STRBP → EIF2A... \n", "10 HES1 TYK2|PROTEIN HES1 → EGR1 → RELA → HRH1 → GRK2 → GNB2 → PTAFR \n", "11 ASGR2 TYK2|PROTEIN ASGR2 → EGR1 → RELA → HRH1 → GRK2 → GNB2 → PTAFR \n", "12 HES1 TYK2|PROTEIN HES1 → TCF4 → CSNK2A1 → THAP1 → STRBP → EIF2AK... \n", "13 IFITM3 TYK2|PROTEIN IFITM3 → STAT1 → TYK2 \n", "14 IL1R1 TYK2|PROTEIN IL1R1 → STAT1 → TYK2 \n", "\n", " score coverage path_fraction length \\\n", "0 0.768584 1.0 0.949464 4 \n", "1 0.763776 1.0 0.859208 3 \n", "2 0.526158 1.0 0.957941 7 \n", "3 0.455786 1.0 0.722857 4 \n", "4 0.421086 1.0 0.950878 3 \n", "5 0.352654 1.0 0.853802 5 \n", "6 0.336833 1.0 0.695527 7 \n", "7 0.313256 1.0 0.500629 3 \n", "8 0.296286 1.0 0.473508 3 \n", "9 0.272955 1.0 0.510312 7 \n", "10 0.217661 1.0 0.519803 7 \n", "11 0.168113 1.0 0.266620 7 \n", "12 0.143988 1.0 0.343862 7 \n", "13 0.119006 1.0 0.133876 3 \n", "14 0.113968 1.0 0.213073 3 \n", "\n", " nodes path_probability \\\n", "0 [IL15|RNA, IRF1|PROTEIN, NPM1|PROTEIN, TYK2|PR... 0.007545 \n", "1 [IFITM3|RNA, STAT2|PROTEIN, TYK2|PROTEIN] 0.008616 \n", "2 [ZC3H12A|RNA, RELA|PROTEIN, HRH1|PROTEIN, GRK2... 0.002227 \n", "3 [ASGR2|RNA, IRF1|PROTEIN, NPM1|PROTEIN, TYK2|P... 0.002566 \n", "4 [RASGEF1B|RNA, STAT1|PROTEIN, TYK2|PROTEIN] 0.001781 \n", "5 [INSIG1|RNA, CCNT2|PROTEIN, HEXIM1|PROTEIN, NP... 0.001408 \n", "6 [KIF2A|RNA, TCF4|PROTEIN, CSNK2A1|PROTEIN, THA... 0.000612 \n", "7 [LAP3|RNA, STAT1|PROTEIN, TYK2|PROTEIN] 0.001983 \n", "8 [LAP3|RNA, STAT2|PROTEIN, TYK2|PROTEIN] 0.001876 \n", "9 [IL1R1|RNA, TCF4|PROTEIN, CSNK2A1|PROTEIN, THA... 0.000580 \n", "10 [HES1|RNA, EGR1|PROTEIN, RELA|PROTEIN, HRH1|PR... 0.000327 \n", "11 [ASGR2|RNA, EGR1|PROTEIN, RELA|PROTEIN, HRH1|P... 0.000947 \n", "12 [HES1|RNA, TCF4|PROTEIN, CSNK2A1|PROTEIN, THAP... 0.000216 \n", "13 [IFITM3|RNA, STAT1|PROTEIN, TYK2|PROTEIN] 0.001343 \n", "14 [IL1R1|RNA, STAT1|PROTEIN, TYK2|PROTEIN] 0.000242 \n", "\n", " score_deg_given_target \n", "0 0.809492 \n", "1 0.888930 \n", "2 0.549259 \n", "3 0.630534 \n", "4 0.442840 \n", "5 0.413040 \n", "6 0.484285 \n", "7 0.625725 \n", "8 0.625725 \n", "9 0.534878 \n", "10 0.418738 \n", "11 0.630534 \n", "12 0.418738 \n", "13 0.888930 \n", "14 0.534878 " ] }, "execution_count": 12, "metadata": {}, "output_type": "execute_result" } ], "source": [ "causal_paths = tc.get_causal_paths(prediction_id=best_prediction_id)\n", "causal_paths.sort_values(by=\"score\", ascending=False).head(15)" ] }, { "cell_type": "markdown", "id": "aa558a8a", "metadata": {}, "source": [ "How much of the psoriasis signature does the target actually reach? `get_all_degs` lists the\n", "DEGs that mapped onto the interactome, and `get_degs_impacted_by_target` lists those with\n", "causal support above the model threshold." ] }, { "cell_type": "code", "execution_count": 13, "id": "e424ba70", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "119 disease DEGs mapped onto the interactome\n", "10 are causally impacted by TYK2|PROTEIN (8% of mapped DEGs)\n", "Examples: ['IFITM3', 'IL15', 'ASGR2', 'LAP3', 'ZC3H12A', 'IL1R1', 'KIF2A', 'RASGEF1B', 'HES1', 'INSIG1']\n" ] } ], "source": [ "all_degs = tc.get_all_degs(prediction_id=best_prediction_id)\n", "impacted_degs = tc.get_degs_impacted_by_target(prediction_id=best_prediction_id)\n", "\n", "print(f\"{len(all_degs)} disease DEGs mapped onto the interactome\")\n", "print(\n", " f\"{len(impacted_degs)} are causally impacted by {best_target} \"\n", " f\"({len(impacted_degs) / max(len(all_degs), 1):.0%} of mapped DEGs)\"\n", ")\n", "print(\"Examples:\", impacted_degs[:12])" ] }, { "cell_type": "markdown", "id": "3d183c9c", "metadata": {}, "source": [ "## Conclusion\n", "\n", "Putting it together to answer the question — *Is Deucravacitinib an interesting drug for psoriasis?*\n", "\n", "- We derived a **psoriasis signature** from patient **dendritic cells** (psoriasis vs. normal).\n", "- We validated Deucravacitinib's target **TYK2** against that signature with TwinCell.\n", "- The **target-validation score** and the **fraction of impacted DEGs** quantify how well\n", " inhibiting TYK2 explains the disease state, and the **causal graph / paths** show the\n", " JAK–STAT routes connecting the disease genes to the target.\n", "\n", "A high score with mechanistically coherent paths (through STAT factors) is supporting evidence\n", "that **Deucravacitinib's mechanism is relevant to psoriasis** — consistent with its approval\n", "for plaque psoriasis.\n", "\n", "**Next steps to try:**\n", "\n", "- Repeat for other cell types (e.g. `helper T cell`, `suprabasal keratinocyte`) to see where the mechanism is strongest.\n", "- Compare against a **negative-control target** (an unrelated protein) on the same signature —\n", " it should score near zero, confirming specificity.\n", "- Validate other psoriasis drugs by swapping in their targets (e.g. `IL17A`, `IL23A`, `JAK1`)." ] }, { "cell_type": "markdown", "id": "65a97176", "metadata": {}, "source": [] } ], "metadata": { "kernelspec": { "display_name": ".venv", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.12.11" } }, "nbformat": 4, "nbformat_minor": 5 }