Perturbome Synergy: An Integrated Web Server for Discovering Synergistic
Drug and Target Combinations Using Perturbational Transcriptomics

CRISPR KO

Drug Perturbation

Transcriptomics

Synergy Pair

Upload Signature
➽
Select Perturbation Library
➽
Run Synergy Scoring
➽
Visualize Results
➽
Download Export

Step 1: Upload Disease Signature

Step 2: Upload Reference Signature

Step 3: Select Small Molecule Dataset


Download Plot as PDF
Download Plot as PDF
Download Table Data

Step 1: Upload Disease Signature

Step 2: Upload Reference Signature

Step 3: Select KO dataset


Download Table Data

1 Motivation

Perturbome Synergy is a web platform for systematic discovery of synergistic drug and target combinations. The efficacy of single-agent oncology drugs is often limited by intrinsic and acquired drug resistance, typically driven by the activation of compensatory signaling pathways in complex cellular networks. Consequently, synergistic drug combinations that target multiple pathways have become a key strategy to enhance antitumor activity, overcome resistance, and improve patient outcomes. Large-scale perturbational transcriptomics projects, such as the LINCS L1000 program, have generated a vast public resource with over 943,000 chemical and 137,000 CRISPR knockout&knockdown gene expression signatures, offering valuable data for the in silico prediction of synergistic interactions. However, significant challenges remain in translating these fragmented and heterogeneous datasets into actionable therapeutic hypotheses. Existing tools lack an integrated, user-friendly framework that allows researchers to directly utilize disease-specific gene signatures and reference perturbations to mine these databases for synergistic partners. To address this, we developed Perturbome Synergy—an accessible web platform designed for researchers without specialized bioinformatics expertise. The platform allows users to: (1) upload custom disease and reference perturbation signatures; (2) query and compute against a consolidated library of small molecules and CRISPR knockouts; (3) perform synergy scoring to rank drug-drug or drug-target combinations; and (4) visualize results via interactive networks and detailed annotations. This approach directly links user data with comprehensive perturbational databases, providing a straightforward and effective framework for discovering synergistic treatment strategies tailored to specific biological and clinical contexts.

Inspired by the SynergySeq platform (http://synergyseq.com/) developed by Vasileios Stathias and colleagues (Nature Communications, 9:5315)

2 Overview

Perturbome Synergy incorporates a total of 1080000 gene sets across 272 cell lines in 21 tissue types and includes four main modules: Synergy Drugs,Synergy Target.

2.1 Data

3 Analysis Workflows

Perturbome Synergy features two analytical modules that utilize chemical and genetic perturbation data from the LINCS L1000 and LINCS 2020 programs to analyze synergistic drug interactions, respectively.

3.1 Synergy Drugs

The Synergy Drugs module in Perturbome Synergy is designed to enable the systematic prediction and analysis of synergistic drug combinations by utilizing large-scale perturbation datasets, such as the LINCS L1000 and LINCS2020 programs. This module allows researchers to assess the combined effects of drug pairs, identify potential synergistic interactions, and prioritize drug combinations for further investigation.

Follow these steps:




① Upload Disease Siganture : a text file describing the differentially expressed genes of the disease state ( such as a certain cancer ). It needs to contain two columns, Gene and Direction ( up or down ).

② Upload Reference Siganture : A document describing the genetic perturbation of the ' reference drug ' that you are concerned about. The document needs to list the genes and their expression status.

③ Users are required to select a dataset for analysis, which integrates over 943,000 gene expression signatures generated by chemical perturbations from large-scale projects such as LINCS L1000. Depending on their specific needs, users can choose between the 'consensus' dataset, which provides aggregated data, or the more granular 'detailed' dataset. The latter includes additional layers of information, such as cell line, perturbation time, and drug dose, offering greater resolution for in-depth analysis.

④ Click ' Submit ' to view the visualization results, scores and rankings of drug synergy in the selected data set in ③.





Upon successful completion of the analysis, users will receive visual representations including a scatter plot of the synergy index and a synergy score plot(Figure 2).The output results allow users to adjust the number of top-ranked drugs displayed in the graph(Figure 2A), and the Quick Tips(Figure 2B) provide guidance on how to interpret the Figure 3C and Figure 3D graphs, thereby facilitating the identification of synergistic drug combinations.We also present a multidimensional view that illustrates the distribution of directional data uploaded by users, specifically focusing on landmark genes and the mechanisms of action (MOA) of FDA-approved drugs. When the Consensus data mode is selected, the display emphasizes gene expression patterns and the distribution of drug mechanisms of action. In contrast, selecting the Detail mode provides a more granular view, highlighting gene signatures, cell lines, and the number of drugs within a specific dataset(Figure 3).




Once an instance is submitted, a results table is generated, sorted by the signature exhibiting the highest degree of synergy. Users can filter the results to identify Drug Repurposing Hubs and FDA-approved drugs. In Detail mode, additional filtering options are available, allowing users to refine results by cell lines, drug concentrations, and time points(Figure 4).

3.2 Synergy Target

The Synergy Target module in Perturbome Synergy identifies synergistic target inhibitors through gene knockout and knockdown signatures or target inhibitor input. Synergistic target inhibitors are identified by uploading gene expression signatures from gene knockout signatures or target inhibitor treatments, which reflect the biological effects of gene inhibition. By comparing these signatures with chemical inhibitor data, the platform identifies synergistic drug combinations that may enhance therapeutic efficacy or overcome resistance mechanisms. The system calculates synergy scores, with higher scores indicating greater therapeutic effects from combined target inhibition, suggesting promising drug combinations for further validation. Additionally, the platform allows exploration of synergy across different cell lines, concentrations, and treatment durations, providing a comprehensive approach to discovering novel therapeutic strategies.

The process here is nearly identical to section 3.1, simply follow these steps:




① Upload Disease Siganture : a text file describing the differentially expressed genes of the disease state ( such as a certain cancer ). It needs to contain two columns, Gene and Direction ( up or down ).

② Upload Reference Siganture : A document describing the genetic perturbation(knockout&knockdown) of the ' Reference ' that you are concerned about. The document needs to list the genes and their expression status.

③ Users are required to select a dataset for analysis, which integrates over 137000 gene expression signatures generated by genetics perturbations from large-scale projects such as LINCS L1000. Depending on their specific needs, users can choose between the 'consensus' dataset, which provides aggregated data, or the more granular 'detailed' dataset. The latter includes additional layers of information, such as cell line, offering greater resolution for in-depth analysis.

④ Click ' Submit ' to view the visualization results, scores and rankings of target synergy in the selected data set in ③.





Upon successful completion of the analysis, users will receive visual representations including a scatter plot of the synergy index and a synergy score plot (Figure 6). The output allows for the adjustment of the number of top-ranked targets displayed in the figure (Figure 6A). Additionally, Quick Tips (Figure 6B) offer guidance on interpreting the graphics in Figures 3C and 3D, which aids in identifying synergistic target-inhibitor combinations. The multidimensional view presented illustrates the distribution of directional data uploaded by the user, as well as the classification of targets. When the 'Consensus' data mode is selected, the display emphasizes the classification of gene expression patterns and targets. Conversely, the 'Detail' mode provides a more granular view, highlighting the number of gene signatures, cell lines, and knockout/knockdown targets within a specific dataset (Figure 7).

What can we do using Perturbome Synergy ?

Perturbome Synergy identifies synergistic drug combinations to enhance anti-tumor efficacy and overcome resistance.

The therapeutic efficacy of a single agent is often limited by its specificity toward a particular molecular target or signaling pathway, whereas the initiation and progression of complex diseases such as cancer are sustained by multiple, interconnected regulatory networks. Identifying synergistic drug combinations enables multi-target perturbation within these networks, thereby amplifying anti-tumor activity and mitigating the emergence of drug resistance. Inhibition of a single pathway frequently triggers compensatory feedback activation or cross-talk among parallel signaling cascades, ultimately diminishing treatment efficacy. Synergistic drug combinations can overcome these adaptive responses by simultaneously blocking both primary and compensatory pathways—such as the PI3K/AKT/mTOR and RAS/RAF/MEK/ERK axes—thus restoring drug sensitivity. Moreover, synergistic screening facilitates the discovery of synthetic lethality interactions, in which inhibition of one pathway becomes lethal only in the presence of specific genetic alterations. Such combinations provide a rational framework for precision oncology, offering new strategies to exploit tumor-specific vulnerabilities.

Here, we demonstrate the utility of Perturbome Synergy through two cases.

Case 1 Searching for drugs synergistic with Lapatinib from LINCS L1000 data

Lapatinib is an FDA-approved dual-target epidermal growth factor receptor 1/2 (EGFR1/2) tyrosine kinase inhibitor indicated for the treatment of HER2-positive advanced or metastatic breast cancer. However, the development of resistance to lapatinib is common over time, with key mechanisms including the activation of alternative signaling pathways such as PI3K/AKT and MAPK, enabling cancer cells to bypass HER2-mediated inhibition. The LINCS L1000 consensus small molecule dataset comprises chemical perturbation profiles for 20,574 small molecules. To address the challenge of drug resistance, we leverage the LINCS L1000 data, incorporating the chemical perturbation signature of lapatinib alongside breast cancer differential analysis. Using the Perturbome Synergy tool, we aim to identify small molecules that exhibit potential synergy with lapatinib for enhancing its therapeutic efficacy in breast cancer treatment.




The ranked list highlighted mechanistic clusters including PI3K/AKT/mTOR pathway inhibitors and HDAC inhibitors1. Among FDA/clinically relevant candidates, MK-2206 (AKT inhibitor) has been evaluated in combination with lapatinib in a phase I study with expansion in advanced HER2+ breast cancer2.

We also employed lapatinib to evaluate the performance of the model in predicting drug combinations. Specifically, we compared the predicted drug combination results for lapatinib in Perturbome Synergy’s LINCS L1000 and LINCS2020 datasets with two external datasets: Drugcomb3 and DrugcombDB4. In the Drugcomb dataset, positive case identification is based on the SynergyZIP threshold (> 10), as outlined in the model paper, while DrugcombDB uses the model’s predefined co-judging labels. ROC curves were generated to quantify the model’s performance, and metrics such as accuracy, precision, recall, and F1 score were computed to assess the model’s effectiveness.

Searching for drugs synergistic with Lapatinib from LINCS L1000 detailed chemical perturbation data

The combination of lapatinib with mTOR inhibitors has garnered increasing attention in anticancer therapy, due to the critical role of the mTOR signaling pathways in the initiation and progression of various cancers. Lapatinib primarily targets HER2 (human epidermal growth factor receptor 2) to treat HER2-positive breast cancer. However, inhibition of the HER2 pathway can trigger compensatory activation of parallel signaling pathways, such as the mTOR pathway, which may contribute to drug resistance. By inputting the signature of lapatinib and the results of BRCA differential analysis into Perturbome Synergy, we focused on analyzing the chemical perturbation profiles related to breast cancer. The results revealed that the top ten drugs identified were predominantly mTOR and HDAC inhibitors.



Case 2 Using an Olaparib (PARP inhibitor) perturbational signature as a query, the model prioritizes the PI3K/AKT/mTOR–DDR (ATR) combination axis in a breast lineage context.

Olaparib is an FDA-approved PARP1/2 inhibitor used in HRD/BRCA-associated breast cancer; however, adaptive resistance frequently develops through rewiring of DNA damage response and replication stress pathways. To systematically nominate actionable co-vulnerabilities, we performed a target-centric synergy analysis rather than a drug-combination screen. Leveraging the LINCS 2020 data, we integrated the chemical perturbation signature of olaparib with breast cancer differential expression profiles, and directly quantified a target-level synergy score for candidate targets, yielding a ranked list of putative co-targets predicted to enhance olaparib efficacy.

To assess model interpretability in pharmacological synthetic lethality/combination settings, we used an olaparib-induced transcriptional perturbation signature as the query and performed reverse retrieval in the LINCS 2020 breast lineage (breast-derived cellular context) dataset. Candidate combination vulnerabilities were interpreted at both the target level and the drug level.



At the target level, the model ranked 1,394 druggable targets, placing ATR at rank 6 (approximately top 0.5%) among the Top10 candidates. The associated “related drug” annotations included multiple ATR inhibitors (AZD6738/ceralasertib, VE-821, AZ20), suggesting that ATR, a central node in replication stress and DNA damage response (DDR) signaling, may represent a key cooperative vulnerability that potentiates PARP inhibition. This signal is consistent with prior studies showing that combined PARP and ATR inhibition amplifies replication stress, promotes replication fork stalling/collapse and genome instability, and yields synergistic cytotoxicity5.



At the drug level, the model returned Top10 candidate synergistic perturbagens from 69,195 MOA/target-annotated drug perturbation records, revealing a clear mechanistic clustering: PI3K–AKT–mTOR pathway inhibitors dominated the Top10, including AZD-5363/capivasertib (AKT inhibitor), AZD-8055 (mTOR inhibitor), BGT-226 (PI3K inhibitor; two independent entries in the Top10), and GDC-0980 (PI3K/mTOR inhibitor). This pattern aligns closely with existing evidence: in breast cancer—particularly TNBC—PI3K inhibition can downregulate BRCA1/2 and impair homologous recombination (HR), thereby sensitizing HR-proficient tumors to PARP inhibition; additionally, mTOR inhibition has been reported to suppress HR and synergize with PARP inhibition, jointly supporting PARP–PI3K/AKT/mTOR as a pharmacological synthetic-lethal/strong-synergy axis6-8.

Furthermore, the olaparib + capivasertib combination has been evaluated in phase I and expansion studies with observable antitumor activity and feasible safety, providing external translational support for the model’s actionable outputs in a breast-relevant context.9-10.

Taken together, the olaparib case simultaneously exhibits drug-level clustering of PI3K/AKT/mTOR inhibitors and target-level prioritization of ATR, forming a coherent, mutually reinforcing loop from signature → synergistic drugs → cooperative targets/pathways. This provides a mechanistically consistent and pharmacologically actionable example demonstrating the model’s ability to capture PARP-centered combination vulnerabilities.

Reference

1 Clayton NS, Carter EP, Fearon AE, et al. HDAC Inhibition Restores Response to HER2-Targeted Therapy in Breast Cancer via PHLDA1 Induction. International Journal of Molecular Sciences. 2023; 24(7):6228.

2 Kari B. Wisinski, Amye J. Tevaarwerk, Mark E. Burkard, et al. Phase I Study of an AKT Inhibitor (MK-2206) Combined with Lapatinib in Adult Solid Tumors Followed by Dose Expansion in Advanced HER2+ Breast Cancer. Clin Cancer Res 1 June 2016; 22 (11): 2659–2667.

3 Bulat Zagidullin, Jehad Aldahdooh, Shuyu Zheng, et al. DrugComb: an integrative cancer drug combination data portal, Nucleic Acids Research, Volume 47, Issue W1, 02 July 2019, Pages W43–W51.

4 Hui Liu, Wenhao Zhang, Bo Zou, et al. DrugCombDB: a comprehensive database of drug combinations toward the discovery of combinatorial therapy, Nucleic Acids Research, Volume 48, Issue D1, 08 January 2020, Pages D871–D881.

5 Lloyd, R.L., Wijnhoven, P.W.G., Ramos-Montoya, A., et al. Combined PARP and ATR inhibition potentiates genome instability and cell death in ATM-deficient cancer cells. Oncogene 39, 4869–4883 (2020).

6 Ibrahim YH, García-García C, Serra V, et al. PI3K inhibition impairs BRCA1/2 expression and sensitizes BRCA-proficient triple-negative breast cancer to PARP inhibition. Cancer Discov. 2012 Nov;2(11):1036-47. doi: 10.1158/2159-8290.CD-11-0348. Epub 2012 Aug 22. PMID: 22915752; PMCID: PMC5125254.

7 Mo W, Liu Q, Lin CC, et al. mTOR Inhibitors Suppress Homologous Recombination Repair and Synergize with PARP Inhibitors via Regulating SUV39H1 in BRCA-Proficient Triple-Negative Breast Cancer. Clin Cancer Res. 2016 Apr 1;22(7):1699-712. doi: 10.1158/1078-0432.CCR-15-1772. Epub 2015 Nov 6. PMID: 26546619; PMCID: PMC4858320.

8 Wei Mo, Qingxin Liu, Curtis Chun-Jen Lin, et al. Mills, Kaiyi Li, Shiaw-Yih Lin; mTOR Inhibitors Suppress Homologous Recombination Repair and Synergize with PARP Inhibitors via Regulating SUV39H1 in BRCA-Proficient Triple-Negative Breast Cancer. Clin Cancer Res 1 April 2016; 22 (7): 1699–1712.

9 Timothy A. Yap, Rebecca Kristeleit, Vasiliki Michalarea, et al. Phase I Trial of the PARP Inhibitor Olaparib and AKT Inhibitor Capivasertib in Patients with BRCA1/2- and Non–BRCA1/2-Mutant Cancers. Cancer Discov 1 October 2020; 10 (10): 1528–1543.

10 Westin SN, Labrie M, Litton JK, et al. Phase Ib Dose Expansion and Translational Analyses of Olaparib in Combination with Capivasertib in Recurrent Endometrial, Triple-Negative Breast, and Ovarian Cancer. Clin Cancer Res. 2021 Dec 1;27(23):6354-6365. doi: 10.1158/1078-0432.CCR-21-1656. Epub 2021 Sep 13. PMID: 34518313; PMCID: PMC8639651.