Tissue-Aware Prototype Learning Model for Predicting Anticancer Drug Response in Patients

Despite large-scale efforts to measure the effect of drug screens in cancer cell lines, mapping the effects of drugs to patient samples has been a challenge. Biological differences between cell lines and patients, such as lack of immune system or microbiome, in-vitro survival adaptations, and biases in measurement technologies create differences across sample modalities that can confound analysis including prediction with machine learning. In this work, we propose a multiway batch correction strategy to enable algorithmic prediction of tumor drug response across model systems and patient data.Recent advances in batch correction algorithms have been motivated by the need to correct for batch effects in single-cell omics and include diverse approaches such as variational autoencoders (VAEs) and generative adversarial networks (GANs). Given the successes of these generative deep learning methods in single cell sequencing analysis, we worked to employ similar approaches to correct large omics measurements across various cancer datasets. Here, we describe mapping of datasets from diverse data sources and model systems to the same space, so that a predictive model of drug response built in a system such as cell lines can be used in biologically relevant models such as organoids, patient derived xenografts, and tumor data. Specifically, we introduce a modified loss function in a VAE using cosine similarity distance to minimize the effect of different cancer model systems in predicting cancer types. We evaluate the method on standard data types for drug response prediction - gene expression, copy number variation, and protein abundance. For this method, the cosine similarity is added as an additional term to the VAE reconstruction and Kullback-Leibler divergence loss terms. This injects a quantification of the dissimilarity between the tumor and tumor model distributions into the backpropagation and gradient descent for updating the model parameters resulting in an encoded representation of the data where the effect of data source has been attenuated while preserving the phenotypic signal. We evaluate our approach for biological signal preservation while reducing model system-specific noise with logistic regression and Euclidean distance. Our results show that the proposed VAE can effectively correct for platform effects and improve the accuracy of downstream integrative analyses. This study has the potential to improve the accuracy and translatability of proteogenomic drug response studies. The proposed modified VAE could be used to correct for platform effects in a variety of datasets, including those from different studies, different platforms, and different cancer types. This could lead to new insights into cancer biology, calibration of cancer patient digital twins, and the development of new diagnostic and therapeutic strategies. Citation Format: Brian Karlberg, Raphael Kirchgaessner, Jeremy R. Jacobson, Kyle Ellrott, Sara J. Gosline. Tumor model to tumor treatment: Applying deep learning approaches to map multimodal data from cancer model systems to patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7393.

MotivationIndividualized drug response prediction is a fundamental part of personalized medicine for cancer. Great effort has been made to discover biomarkers or to develop machine learning methods for accurate drug response prediction in cancers. Incorporating prior knowledge of biological systems into these methods is a promising avenue to improve prediction performance. High-throughput cell line assays of drug-induced transcriptomic perturbation effects are a prior knowledge that has not been fully incorporated into a drug response prediction model yet.ResultsWe introduce a unified probabilistic approach, Drug Response Variational Autoencoder (Dr.VAE), that simultaneously models both drug response in terms of viability and transcriptomic perturbations. Dr.VAE is a deep generative model based on variational autoencoders. Our experimental results showed Dr.VAE to do as well or outperform standard classification methods for 23 out of 26 tested Food and Drug Administration-approved drugs. In a series of ablation experiments we showed that the observed improvement of Dr.VAE can be credited to the incorporation of drug-induced perturbation effects with joint modeling of treatment sensitivity.Availability and implementationProcessed data and software implementation using PyTorch (Paszke et al., 2017) are available at: https://github.com/rampasek/DrVAE.Supplementary informationSupplementary data are available at Bioinformatics online.

Drug response prediction (DRP) is important for precision medicine to predict how a patient would react to a drug before administration. Existing studies take the cell line transcriptome data, and the chemical structure of drugs as input and predict drug response as IC50 or AUC values. Intuitively, use of drug target interaction (DTI) information can be useful for DRP. However, use of DTI is difficult because existing drug response database such as CCLE and GDSC do not have information about transcriptome after drug treatment. Although transcriptome after drug treatment is not available, if we can compute the perturbation effects by the pharmacologic modulation of target gene, we can utilize the DTI information in CCLE and GDSC. In this study, we proposed a framework that can improve existing deep learning-based DRP models by effectively utilizing drug target information. Our framework includes NetGP, a module to compute gene perturbation scores by the network propagation technique on a network. NetGP produces genes in a ranked list in terms of gene perturbation scores and the ranked genes are input to a multi-layer perceptron to generate a fixed dimension vector for the integration with existing DRP models. This integration is done in a model-agnostic way so that any existing DRP tool can be incorporated. As a result, our framework boosts the performance of existing DRP models, in 64 of 72 comparisons. The performance gains are larger especially for test scenarios with samples with unseen drugs by large margins up to 34% in Pearson's correlation coefficient.

One of the most difficult challenges in precision medicine is determining the best treatment strategy for each patient based on personal information. Since drug response prediction in vitro is extremely expensive, time-consuming and virtually impossible, and because there are so many cell lines and drug data, computational methods are needed. MinDrug is a method for predicting anti-cancer drug response which try to identify the best subset of drugs that are the most similar to other drugs. MinDrug predicts the anti-cancer drug response on a new cell line using information from drugs in this subset and their connections to other drugs. MinDrug employs a heuristic star algorithm to identify an optimal subset of drugs and a regression technique known as Elastic-Net approaches to predict anti-cancer drug response in a new cell line. To test MinDrug, we use both statistical and biological methods to assess the selected drugs. MinDrug is also compared to four state-of-the-art approaches using various k-fold cross-validations on two large public datasets: GDSC and CCLE. MinDrug outperforms the other approaches in terms of precision, robustness and speed. Furthermore, we compare the evaluation results of all the approaches with an external dataset with a statistical distribution that is not exactly the same as the training data. The results show that MinDrug continues to outperform the other approaches. MinDrug's source code can be found at https://github.com/yassaee/MinDrug. Supplementary data are available at Bioinformatics online.

Drug repurposing is the process of identifying new clinical indications for an existing drug. Some of the recent studies utilized drug response prediction models to identify drugs that can be repurposed. By representing cell-line features as a pathway-pathway interaction network, we can better understand the connections between cellular processes and drug response mechanisms. Existing deep learning models for drug response prediction do not integrate known biological pathway-pathway interactions into the model. This paper presents a drug response prediction model that applies a graph convolution operation on a pathway-pathway interaction network to represent features of cancer cell-lines effectively. The model is used to identify potential drug repurposing candidates for Non-Small Cell Lung Cancer (NSCLC). Experiment results show that the inclusion of graph convolutional model applied on a pathway-pathway interaction network makes the proposed model more effective in predicting drug response than the state-of-the-art methods. Specifically, the model has shown better performance in terms of Root Mean Squared Error, Coefficient of Determination, and Pearson's Correlation Coefficient when applied to the GDSC1000 dataset. Also, most of the drugs that the model predicted as top candidates for NSCLC treatment are either undergoing clinical studies or have some evidence in the PubMed literature database.

Accurate prediction of patient drug response is required for successful personalized medicine. Cancer is a disease resulting from myriad genetic alterations. Accordingly, it should be possible to predict drug response of specific cancers using genetic and molecular signatures. To aid this effort, an ongoing collaboration established between Novartis and the Broad Institute called the Cancer Cell Line Encyclopedia (CCLE) has (i) comprehensively characterized genome-scale mRNA expression, copy number alteration and mutation profiles for nearly 1000 cancer cell line models spanning many tumor types and (ii) profiled ~500 of these cell lines against ~1000 anticancer compounds. Using these data, we are developing a scalable, extensible predictive modeling framework based on various supervised learning methods that include both categorical classification and linear regression approaches to predict compound sensitivity. By cross-validation of the generated models against test data sets, we quantitatively estimated model performance. Our initial results validate several preexisting genetic predictors of sensitivity. In addition, our models reported 70% or higher sensitivity and specificity for a number of compounds and yielded interesting potentially novel predictive features that, in some cases, outperform previously existing genetic predictors of sensitivity. Our integrative approach demonstrates that pharmacological profiling of large, genomically annotated cancer model systems, coupled with systematic predictive modeling, can uncover novel predictors of drug response and may ultimately aid patient stratification. This talk is also presented as Poster A4.

Background and objectives: Breast cancer is the most prevalent type of cancer among women. The effectiveness of anticancer pharmacological therapy may get adversely affected by tumor heterogeneity that includes genetic and transcriptomic features. This leads to clinical variability in patient response to therapeutic drugs. Anticancer drug design and cancer understanding require precise identification of cancer drug responses. The performance of drug response prediction models can be improved by integrating multi-omics data and drug structure data. Methods: In this paper, we propose an Autoencoder (AE) and Graph Convolutional Network (AGCN) for drug response prediction, which integrates multi-omics data and drug structure data. Specifically, we first converted the high dimensional representation of each omic data to a lower dimensional representation using an AE for each omic data set. Subsequently, these individual features are combined with drug structure data obtained using a Graph Convolutional Network and given to a Convolutional Neural Network to calculate IC[Formula: see text] values for every combination of cell lines and drugs. Then a threshold IC[Formula: see text] value is obtained for each drug by performing K-means clustering of their known IC[Formula: see text] values. Finally, with the help of this threshold value, cell lines are classified as either sensitive or resistant to each drug. Results: Experimental results indicate that AGCN has an accuracy of 0.82 and performs better than many existing methods. In addition to that, we have done external validation of AGCN using data taken from The Cancer Genome Atlas (TCGA) clinical database, and we got an accuracy of 0.91. Conclusion: According to the results obtained, concatenating multi-omics data with drug structure data using AGCN for drug response prediction tasks greatly improves the accuracy of the prediction task.

Accurate prediction of drug response using machine learning (ML) remains a significant challenge in drug development as well as in personalized cancer therapy. Furthermore, there is a scarcity of rigorous external validation studies for evaluating drug response prediction models. Even when available, such validations are often confined to in-vitro studies. Orthotopic patient-derived xenograft (O-PDX) mice serve as essential preclinical models that closely replicate the human tumor microenvironment and assess the therapeutic response to cancer treatments. We present CertisAITM, a novel ensemble of machine learning models that have been trained on a vast array of experimental high-throughput screenings of both monotherapies and combination therapies, incorporating over 4500 investigational and FDA-approved drugs across 10 major cancer indications. The results of our ML drug response prediction models show an average R2 of 0.75 and RMSE of 0.62 (across internal 5-fold cross-validation) for all 10 cancer prediction models. Our evaluation of CertisAI's drug response predictions within O-PDX pharmacology studies has shown a notable overall correlation between actual observed tumor growth inhibition (TGI) and predicted TGI, with an r value of 0.45 across seven indications. The most accurate prediction model achieved an r value of 0.7 for colorectal cancer, encompassing 37 treatments in 6 O-PDX studies. Altogether, these results highlight CertisAI’s potential in enhancing pre-clinical model selection as well as personalized treatment strategies in oncology. Citation Format: Yuan-Hung Chien, Raffaella Pippa, Warren Andrews, Javier Rodriguez, Kristen Buck, Derrick Gorospe, Elizabeth Valencia, Bianca Carapia, Emily Eastwood, Jantzen Sperry, Jonathan Nakashima, Long Do. Assessing machine learning models of drug response predictions using orthotopic patient-derived xenograft [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3537.

BackgroundOne of the challenges in personalized medicine is to determine specific drugs and their dosages for patient individuals who are undergoing a common disease. The technique of cell lines provides a safe approach to capture the drug responses of patient individuals when given specific drugs with varied dosages. However, it is still costly to determine drug responses in cells w.r.t dosages by biological assays. Computational methods provide a promising screening to infer possible drug responses in the cells of patient individuals on a large scale. Nevertheless, existing computational approaches are insufficient to interpret the underlying reason for drug responses.MethodsIn this work, we propose an interpretable model for analyzing and predicting drug responses across cell lines. The proposed model bridges drug features ( e.g., chemical structure fingerprints), cell features ( e.g., gene expression profiles), and drug responses across cells (measured by IC50) by a triple matrix factorization (TMF), such that the underlying reason for drug responses in specific cells is possibly interpreted.ResultsThe comparison with state‐of‐the‐art computational approaches demonstrates the superiority of our TMF. More importantly, a case study of drug responses in lung‐related cell lines shows its interpretable ability to find out highly occurring drug substructures, crucial mutated genes, as well as significant pairs between substructures and mutated genes in terms of drug sensitivity and resistance.ConclusionTMF is an effective and interpretable approach for predicting cell lines responses to drugs, and can dig out crucial pairs of chemical substructures and genes, which uncovers the underlying reason for drug responses in specific cells.

Preclinical cancer model systems, such as large-scale drug screens in cancer cell lines, are pivotal for identifying drug-sensitive indications and biomarkers of drug efficacy. However, these systems often lack the tumor microenvironment's complexity, including interactions with immune cells, which play a critical role in modulating drug response. Investigating drug-induced effects on immune cells is experimentally challenging due to the diversity of immune cell types and the complexity of co-culture systems. Computational models of immune cells can help to overcome these limitations and to guide further experiments. Leveraging public post-perturbation single-cell RNA sequencing datasets from various immune cells (including CD8 and CD4 T cells), we trained Turbine’s Simulated Cell engine to predict transcription factor (TF) activity changes in 17 immune cell types post-drug treatment. We compared the predicted, drug induced TF activity changes to immune killing effects from public co-culture experiments. Predicted activity changes of CD8 T-cell activator TFs, such as FOSL and MYC, positively correlated with tumor-killing effects (Pearson r = 0.35 and 0.42, respectively). Conversely, inhibitor TFs such as NR4A1 and IRF4 exhibited inverse correlations (Pearson r = -0.27 and -0.47, respectively).To assess clinical relevance, we analyzed oncology clinical trial data from ACTT. We calculated mean overall response rates (ORRs) for indication - drug pairs and compared it to in vitro indication specific relative IC50 values of drugs. Using a linear model to predict ORRs from in vitro cancer cell line drug response data alone, we predicted ORRs with a Pearson correlation of 0.28. Integrating simulated drug induced TF activities of CD8 T-cells improved this correlation to 0.48, showing that combining tumor-intrinsic mechanisms with immune-mediated cell killing enhances the predictive performance of the model. Our findings suggest that augmenting traditional in vitro drug screening with simulated immune cell activity significantly enhances the clinical translatability of preclinical drug response models, paving the way for improved therapeutic strategies. Citation Format: Gema Sanz, Réka Böröcz, Bence Czakó, Gerold Csendes, Csaba Gór, Gábor Kovács, László Mérő, Kristóf Szalay, Daniel Veres, Bence Szalai. Computational modeling of immune cell phenotypes enhances prediction of clinical anti-cancer drug response [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5803.

Background: Precision medicine (PM) has transformed cancer treatment but remains effective for only a subset of patients. Most tumors profiled with comprehensive genomic testing lack actionable biomarkers, and even when present, many patients fail to benefit from corresponding therapies. Some without recognized biomarkers respond in off-label contexts, highlighting the need for better biomarkers to capture drug response biology and deliver on the promise of PM. Osimertinib, a third-generation EGFR inhibitor, is used in conjunction with the single gene biomarker, EGFR. Response rates vary by EGFR variant and off-target activity has been reported. We hypothesized that integrating multi-modal data with machine learning could enhance osimertinib response predictions beyond the single-gene EGFR biomarker. Methods: We developed an AI/ML-enabled drug response prediction model that integrates clinical and molecular features with drug structure and functional data to generate drug response predictions. The model produces an AI-enabled composite biomarker (AI-CB) based on a small amount of clinical and molecular patient input data. Each prediction is paired with a proprietary Vulnerability Network™ (VN), providing insights into biological dependencies underlying response predictions. Osimertinib response predictions were evaluated retrospectively across multiple real-world (rw) NSCLC cohorts, including clinicogenomic data from liquid biopsy. The AI-CB was also validated in EGFR-negative NSCLC patient-derived xenografts (PDX). Results: Across multiple osimertinib-treated EGFR+ rw-NSCLC cohorts, the AI-CB accurately stratified patients by sensitivity to osimertinib, despite the presence of EGFR alterations in all patients. EGFR+ patients predicted to be insensitive showed a significant reduction in rw-overall survival compared to those predicted to be sensitive, indicating that osimertinib sensitivity may be impacted by more than just EGFR alterations. Compared to EGFR+ osimertinib insensitive patients, post-hoc and VN analysis across these two cohorts linked sensitivity to upregulation of pathways downstream of EGFR including NOTCH and TP63 and downregulation of RAS/RAF/MEK activity. The AI-CB was successfully validated in human PDX models without activating EGFR alterations. AI-CB application to AACR Project GENIE cohorts suggest that sensitivity to osimertinib could extend to a novel group of patients with NSCLC and beyond, e.g., colorectal cancer. Conclusion: ML enables the integration of diverse data types relevant to drug response, overcoming limitations of single-dimensional approaches. Incorporating AI-driven patient stratification models into drug development and clinical practice could greatly expand the utility of targeted therapies and create effective stratification strategies for drugs lacking adequate biomarkers. Citation Format: Maayan Baron, Felicia Kuperwaser, Colin Tang, Sunil Kumar, Dillon Tracy, Zong Miao, Olena Marchenko, Sepideh Foroutan, Amy Sheide, Nathaniel Tann, Samantha Pindak, Jacob Kaffey, Jordan Wolinsky, Sean Klei, Andrey Chursov, Brandon Funkhouser, Nick Lee, Patrick Bohan, Fahad Khan, Ripple Khera, Jean Michel Rouly, George Komatsoulis, Jeff Sherman, Emily Vucic, Rachael Brake. Identifying novel drivers of drug sensitivity using an AI-enabled composite biomarker - osimertinib sensitivity beyond EGFR [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 4640.

Simple SummaryPreclinical drug screening experiments for anti-cancer drug discovery typically involve testing candidate drugs against cancer cell lines. This process can be expensive and time consuming since the possible experimental space can be quite huge, involving all of the combinations of candidate cell lines and drugs. Guiding drug screening experiments with active learning strategies could potentially identify promising candidates for successful experimentation. This study investigates various active learning strategies for selecting experiments to generate response data for identifying effective treatments and improving the performance of drug response prediction models. We have demonstrated that most active learning strategies are more efficient than random selection for identifying effective treatments.It is well-known that cancers of the same histology type can respond differently to a treatment. Thus, computational drug response prediction is of paramount importance for both preclinical drug screening studies and clinical treatment design. To build drug response prediction models, treatment response data need to be generated through screening experiments and used as input to train the prediction models. In this study, we investigate various active learning strategies of selecting experiments to generate response data for the purposes of (1) improving the performance of drug response prediction models built on the data and (2) identifying effective treatments. Here, we focus on constructing drug-specific response prediction models for cancer cell lines. Various approaches have been designed and applied to select cell lines for screening, including a random, greedy, uncertainty, diversity, combination of greedy and uncertainty, sampling-based hybrid, and iteration-based hybrid approach. All of these approaches are evaluated and compared using two criteria: (1) the number of identified hits that are selected experiments validated to be responsive, and (2) the performance of the response prediction model trained on the data of selected experiments. The analysis was conducted for 57 drugs and the results show a significant improvement on identifying hits using active learning approaches compared with the random and greedy sampling method. Active learning approaches also show an improvement on response prediction performance for some of the drugs and analysis runs compared with the greedy sampling method.

Molecular subtype is a critical determinant of therapeutic approaches in breast cancer. The subtype is based on the expression of three receptors: The Her2/ErbB2 receptor tyrosine kinase (RTK), whose over-expression defines the HER2amp subtype, and the estrogen or progesterone nuclear hormone receptors, whose over-expression defines the HR+ subtype. Triple negative breast cancers (TNBCs) express low levels of all three receptors. HER2amp status serves as a biomarker for therapy with anti-Her2/ErbB2 antibodies such as trastuzumab or pertuzumab and HR+ status is a biomarker for therapy with hormone receptor antagonists such as tamoxifen. TNBCs are usually treated with cytotoxic chemotherapy. However, breast cancer subtypes are heterogeneous, clinical and mRNA subtypes are not identical, and even the best available biomarker, HER2amp status, correctly predicts response to trastuzumab in only a subset of patients. Projects such as the Cancer Cell Line Encyclopedia aim to identify drug response determinants by relating dose-response profiles for multiple compounds to genomic features across cancer cell collections. Recent studies have suggested that measures of target activity such as phosphorylation are more correlated with drug response than mutation status or gene expression, and steady-state protein data is becoming available for tumors and cell lines. However, few studies have tried to relate the biochemistry of signal transduction to drug response on a large scale, mainly because the measurements are demanding and the utility of the approach is unproven. In this work, we ask whether measurement of the basal and perturbed states of immediate-to-early signaling proteins in serum-starved cells, a robust measurement that can be collected with relatively little biological variability, is predictive of tumor subtype as well as of drug response. We measured the abundance and basal phosphorylation state of nuclear and cell surface receptors and of downstream signaling kinases in the standardized NCI-ICBP43 cell line collection. Because perturbation reveals features of signal transduction that are not apparent by steady state profiling, and because biological ligands present in the microenvironment can alter drug sensitivity, we also collected response profiles by exposing cells to a diverse set of growth factors and cytokines and measuring the activities of downstream signaling kinases before and after ligand addition. We present how the resulting set of ∼3×105 receptor and cell response measurements segregated with clinical subtype and how these proteomic measurements predict the sensitivity of cells to a range of targeted anti-cancer drugs. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P6-05-01.

Data from several large high-throughput drug response screens have become available to the scientific community recently. Although many efforts have been made to use this information to predict drug sensitivity, our ability to accurately predict drug response based on genetic data remains limited. In order to systematically examine how different aspects of modelling affect the resulting prediction accuracy, we built a range of models for seven drugs (erlotinib, pacliatxel, lapatinib, PLX4720, sorafenib, nutlin-3 and nilotinib) using data from the largest available cell line and xenograft drug sensitivity screens. We found that the drug response metric, the choice of the molecular data type and the number of training samples have a substantial impact on prediction accuracy. We also compared the tasks of drug response prediction with tissue type prediction and found that, unlike for drug response, tissue type can be predicted with high accuracy. Furthermore, we assessed our ability to predict drug response in four xenograft cohorts (treated either with erlotinib, gemcitabine or paclitaxel) using models trained on cell line data. We could predict response in an erlotinib-treated cohort with a moderate accuracy (correlation ≈ 0.5), but were unable to correctly predict responses in cohorts treated with gemcitabine or paclitaxel.

PTPRF interacting protein alpha 1 (PPFIA1) encodes for liprin-α1, a member of the leukocyte common antigen-related protein tyrosine phosphatase (LAR-RPTPs)-interacting protein family. Liprin-α1 localizes to adhesive and invasive structures in the periphery of cancer cells, where it modulates migration and invasion in head and neck squamous cell carcinoma (HNSCC) and breast cancer. To study the possible role of liprin-α1 in anticancer drug responses, we screened a library of oncology compounds in cell lines with high endogenous PPFIA1 expression. The compounds with the highest differential responses between high PPFIA1-expressing and silenced cells across cell lines were inhibitors targeting mitogen-activated protein kinase kinase (MEK)/extracellular signal-regulated kinases (ERK) signaling. KRAS proto-oncogene, GTPase (KRAS)-mutated MDA-MB-231 cells were more resistant to trametinib upon PPFIA1 knockdown compared with control cells. In contrast, liprin-α1-depleted HNSCC cells with low RAS activity showed a context-dependent response to MEK/ERK inhibitors. Importantly, we showed that liprin-α1 depletion leads to increased p-ERK1/2 levels in all our studied cell lines independent of KRAS mutational status, suggesting a role of liprin-α1 in the regulation of MAPK oncogenic signaling. Furthermore, liprin-α1 depletion led to more pronounced redistribution of RAS proteins to the cell membrane. Our data suggest that liprin-α1 is an important contributor to oncogenic RAS/MAPK signaling, and the status of liprin-α1 may assist in predicting drug responses in cancer cells in a context-dependent manner.