Table · dataset · 2026
Dataset, Model Weights and Code for: PAIRWISE: Deep Learning-based Prediction of Effective Personalized Drug Combinations in Cancer Response
Listed in figshare
<p dir="ltr">PAIRWISE: Deep Learning-based Prediction of Effective Personalized Drug Combinations in CancerPAIRWISE predicts whether a drug pair acts synergistically in a specific tumour sample.
Description
It fuses three modalities — molecular graphs of the two compounds, theirdrug–target interaction profiles propagated over a protein–protein interactionnetwork, and the sample transcriptome — through an attention encoder into a singlesynergy probability.</p><p dir="ltr"><br>This repository contains the model, the seven benchmarked baselines, and the code that reproduces figure and table in the manuscript.
Also you can refer to the code repo in the<a href="github.com/Mew233/pairwise" target="_blank" rel="noreferrer"> GitHub repository</a>. Interactive predictions: <a href="mew233.shinyapps.io/synergyy_shinyr/">synergy explorer</a> and <a href="mew233.shinyapps.io/PAIRWISE_Explorer/">BTKi explorer<br></a></p><h3 dir="ltr">Contents of this deposit</h3><ul><li><code><strong>data/</strong></code> — the p13 benchmark corpus (~30K drug–drug–cell-line records, 1,275 drugs × 163 cell lines across 15 lineages, harmonised from 13 public screens with Loewe/Bliss/ZIP/HSA scores recomputed in SynergyFinder v3.0), plus chemical, drug–target and transcriptome feature banks, the STRING PPI network, and gene sets.</li><li><code><strong>weights/</strong></code> — trained checkpoints for PAIRWISE and eleven baselines (seven published deep-learning methods and four classical ML models). <code>best_model_pairwise.pth</code> is the main model (held-out test AUROC 0.8444).</li><li><code><strong>results/</strong></code> — out-of-fold and held-out predictions for every model.</li><li><code><strong>paper/</strong></code> — per-stage inputs, outputs and scripts behind each figure and table, one directory per analysis (benchmark, wet-lab screen, network/pathway, patient stratification, external DLBCL validation, NCI-DREAM comparison, ablation, SHAP, supplementary).
Read the rest (2 more)
Each has its own README.</li><li><code><strong>example/</strong></code> — a runnable notebook with sample data, sample weights and the original run logs; so you can freely make a prediction under the instruction.</li></ul><h3 dir="ltr">Usage</h3><p dir="ltr"><b>Clone the repository and install:</b></p><pre><pre>git clone github.com/Mew233/pairwise.git && cd pairwise<br>conda create -n pairwise python=3.10 && conda activate pairwise<br>pip install torch==2.3.0 --index-url download.pytorch.org/whl/cu121<br>pip install dgl==2.4.0 -f data.dgl.ai/wheels/torch-2.3/cu121/repo.html<br>pip install -e .</pre></pre><p><br></p><p dir="ltr"><b>Download and place the data.</b> Extract thei zip file you will automatically dump <code>data/</code>, <code>weights/</code> and <code>results/</code> into the repository root, or leave them elsewhere and point the package at them:</p><pre><pre>export PAIRWISE_DATA_ROOT=/path/to/deposit/data<br>export PAIRWISE_WEIGHTS_DIR=/path/to/deposit/weights</pre></pre><p><br></p><p dir="ltr"><b>Run a quick test.</b> Open <code>example/example2run.ipynb</code>, which walks through prediction, feature extraction, fine-tuning and training on the bundled sample data.</p><p dir="ltr">To retrain from scratch (5-fold CV on p13):</p><pre><code>python -m pairwise.main --model pairwise --synergy_df p13 --train_test_mode train</code></pre><p dir="ltr">For environment setup, the full training pipeline and per-analysis instructions, see the <code>README.md</code> in the GitHub repository and the stage READMEs under <code>paper/</code>.</p><h3 dir="ltr">Notes</h3><p dir="ltr">Data files remain subject to the terms of their original sources (DrugComb and the 13 constituent screens, CCLE/DepMap, TCGA, STRING, DrugTargetCommons, DrugBank, the NCI-DREAM Challenge, and Griner et al.).
Code is released under the MIT license.</p><p dir="ltr"><b>Associated Publication:</b> Xu C., et al. "PAIRWISE: Deep Learning-based Prediction of Effective Personalized Drug Combinations in Cancer", under revision (2026).</p><p><br></p>
Links
Where it is published
- DOI doi.org/10.6084/m9.figshare.33950731.v1 ↗
DOI / persistent id · from figshare com
Catalogue records · 1
- OAI-PMH record api.figshare.com/v2/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Af… ↗
metadata API · from figshare com
Topics
Provenance · 1 source records, 19 field assertions
| Source | Key | Last seen | Raw |
|---|---|---|---|
| figshare | oai:figshare.com:article/33950731 | 4 d ago | JSON v1 |
| Field | Assertion | Extractor | Evidence |
|---|---|---|---|
| access_level | source · figshare com | connector:figshare_com@1.0.0 | |
| concepts[disease].local:disease:cancer | enrichment · figshare com | keyword-concept-rules@1.0.0 | title+description (75%) |
| concepts[field].anzsrc:field:321104 | mapping · figshare com | vocabulary-mapper@1.0.0 | keywords['Cancer Therapy'] |
| concepts[field].local:field:astronomy | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:chemistry | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:computer-science-ai | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:earth-environmental | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:economics-finance | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:engineering | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:humanities | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:life-sciences | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:medicine-health | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:ocean-atmospheric | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[field].local:field:social-science | mapping · figshare com | connector:figshare_com@1.0.0 | |
| concepts[modality].local:modality:tabular | enrichment · figshare com | keyword-concept-rules@1.0.0 | title+description (65%) |
| description | source · figshare com | connector:figshare_com@1.0.0 | /metadata/dc/description |
| license | source · figshare com | connector:figshare_com@1.0.0 | /metadata/dc/rights |
| publication_date | source · figshare com | connector:figshare_com@1.0.0 | |
| title | source · figshare com | connector:figshare_com@1.0.0 | /metadata/dc/title |