Predict Drug Response Research Articles

Performance Assessment and Quality Control of Fluorescence Molecular Endoscopy With a Multi-Parametric Rigid Standard.

Fluorescence molecular endoscopy (FME) is emerging as a "red-flag" technique with potential to deliver earlier, faster, and more personalized detection of disease in the gastrointestinal tract, including cancer, and to gain insights into novel drug distribution, dose finding, and response prediction. However, to date, the performance of FME systems is assessed mainly by endoscopists during a procedure, leading to arbitrary, potentially biased, and heavily subjective assessment. This approach significantly affects the repeatability of the procedures and the interpretation or comparison of the acquired data, representing a major bottleneck towards the clinical translation of the technology. Herein, we propose a robust methodology for FME performance assessment and quality control that is based on a novel multi-parametric rigid standard. This standard enables the characterization of an FME system's sensitivity through a single acquisition, performance comparison of multiple systems, and, for the first time, quality control of a system as a function of time and number of usages. We show the photostability of the standard experimentally and demonstrate how it can be used to characterize the performance of an FME system. Moreover, we showcase how the standard can be employed for quality control of a system. In this study, we find that the use of composite fluorescence standards before endoscopic procedures can ensure that an FME system meets the performance criteria and that components prone to performance degradation are replaced in time, avoiding disruption of clinical endoscopy logistics. This will help overcome a major barrier for the translation of FME into the clinics.

3D organoid models for predicting drug response in platinum-sensitive, high-grade serous ovarian cancer

3D organoid models for predicting drug response in platinum-sensitive, high-grade serous ovarian cancer

Integrated mRNA-seq and miRNA-seq analysis reveals key transcription factors of HNF4α and KLF4 in ADPKD

Autosomal Dominant Polycystic Kidney Disease (ADPKD) is the most prevalent genetic disorder affecting the kidneys. Understanding epigenetic regulatory mechanisms and the role of microRNAs (miRNAs) is crucial for developing therapeutic interventions. Two mRNA datasets (GSE7869 and GSE35831) and miRNA expression data (GSE133530) from ADPKD patients were used to find differentially expressed genes (DEGs) and differentially expressed miRNAs (DEMs), with a focus on genes regulated by hub transcription factors (TFs) and their target genes. The expression of hub TFs was validated in human kidneys and animal models through Western Blot (WB) and RT-PCR analysis. The location of the hub TF proteins in kidney cells was observed by a laser confocal microscope. A total of 2037 DEGs were identified. DEM analysis resulted in 59 up-regulated and 107 down-regulated miRNAs. Predicted target DEGs of DEMs indicated two top dysregulated TFs: hepatocyte nuclear factor 4 alpha (HNF4α) and Kruppel-like factor 4 (KLF4). RT-PCR, WB, and immunochemistry results showed that mRNA and protein levels of HNF4α were significantly decreased while KLF4 levels were significantly up-regulated in human ADPKD kidneys and Pkd1 conditional knockout mice compared with normal controls. Laser confocal microscopy revealed that KLF4 was mainly located in the cytoplasm while HNF4α was in the nucleus. Functional enrichment analysis indicated that genes regulated by HNF4α were mainly associated with metabolic pathways, while KLF4-regulated genes were linked to kidney development. Drug response prediction analysis revealed potential drug candidates for ADPKD treatment, including BI-2536, Sepantronium, and AZD5582. This integrated analysis provides new epigenetic insights into the complex miRNA-TF-mRNA network in ADPKD and identifies HNF4α and KLF4 as key TFs. These findings offer valuable resources for further research and potential drug development for ADPKD.

Organoids research progress in gynecological cancers: a bibliometric analysis

Background Gynecological cancers (GC) pose a severe threat to the health and safety of women’s lives, and organoids, as in-vitro research models, have demonstrated significant advantages in simulating tissue characteristics and drug screening. In recent years, there has been a rapid increase in research outcomes related to organoids in GC. However, there has been no bibliometric study concerning.Methods Publications related to GC and organoids from 2010-2023 were retrieved from the Web of Science Core Collection (WoSCC). We conducted a bibliometric analysis and visualization using CiteSpace, VOSviewer, and the Bibliometrix R Package. This analysis included the spatiotemporal distribution, author, sources, references, and keywords.Results A total of 333 publications were included. The number of annual publications indicated an explosive phase of development since 2019. The USA was the most important country in terms of cooperation, publication output, citation and centrality. University of California system ranked first in productivity among institutions, and HIPPO Y is the most relevant author in the research field. CANCERS published the most documents, and NATURE is the most cited sources. Analysis of Keywords and References, it is possible to establish the trend, and find the hotspots in the research field.Conclusion This bibliometric analysis delineated global landscapes and progress trends in GC organoids research. This study emphasized that organoids can effectively replicate the original tissue or tumors, providing a good in-vitro model for research on tumor-related mechanisms and showing significant advantages in drug screening and efficacy clinical prediction. Additionally, as preclinical models, they provide compelling evidence for personalized therapy and prediction of patient drug responses.

OmicsFootPrint: a framework to integrate and interpret multi-omics data using circular images and deep neural networks.

The OmicsFootPrint framework addresses the need for advanced multi-omics data analysis methodologies by transforming data into intuitive two-dimensional circular images and facilitating the interpretation of complex diseases. Utilizing deep neural networks and incorporating the SHapley Additive exPlanations algorithm, the framework enhances model interpretability. Tested with The Cancer Genome Atlas data, OmicsFootPrint effectively classified lung and breast cancer subtypes, achieving high area under the curve (AUC) scores-0.98±0.02 for lung cancer subtype differentiation and 0.83±0.07 for breast cancer PAM50 subtypes, and successfully distinguished between invasive lobular and ductal carcinomas in breast cancer, showcasing its robustness. It also demonstrated notable performance in predicting drug responses in cancer cell lines, with a median AUC of 0.74, surpassing nine existing methods. Furthermore, its effectiveness persists even with reduced training sample sizes. OmicsFootPrint marks an enhancement in multi-omics research, offering a novel, efficient and interpretable approach that contributes to a deeper understanding of disease mechanisms.

Establishment of matched bladder cancer PDX and PDX-derived organoid model and evaluation of anti-tumor efficacy of abemaciclib.

Bladder cancer is one of the most common malignancies of the urinary system and there's a significant unmet need for new effective therapeutics for bladder cancer. The limited number of available models to study malignant bladder tumors is one of the obstructions in developing bladder cancer therapeutics. Patient-derived xenograft (PDX) and organoid (PDO) models are more representatives of human cancer than cell lines and cell line-derived xenograft (CDX) and are likely to be more promising and efficient in predicting drug response and finding new therapeutics. Three pairs of patient-derived xenograft (PDX) models of bladder cancer and their corresponding PDX-derived organoids (PDXOs) were successfully established. These models were utilized to assess the efficacy of abemaciclib. The sensitivity of the drug was determined through the Cell Counting Kit-8 (CCK8) assay in PDXO cultures, corroborated by the EdU incorporation assay. Additionally, the in vivo tumor growth was monitored in the matched PDX models. In vitro PDXO cultures and in vivo PDX tumor models consistently demonstrated that abemaciclib had varying degrees of inhibitory effects across different bladder cancer (BC) patients. Notably, our study further revealed that treatment with abemaciclib significantly modified the expression patterns of CyclinD1/CDK4. This was achieved by not only diminishing their expression levels but also by shifting their expression from a membrane-associated localization to the nucleus. Our research provided compelling evidence attesting to the reliability and potential of PDX and PDXO models in the realm of precision medicine. These models are instrumental in identifying patients who are likely to respond favorably to a specific drug treatment.

Single-cell technology for drug discovery and development

The success rate of drug development today remains low, with long development cycles and high costs, especially in areas such as oncology, neurology, immunology, and infectious diseases. Single-cell omics, encompassing transcriptomics, genomics, epigenomics, proteomics, and metabolomics enable the analysis of gene expression profiles and cellular heterogeneity from the perspective of individual cells, offering a high-resolution view of their functional diversity. These technologies can help reveal disease mechanisms, drug target identification and validation, selection of preclinical models and candidate drugs, and clinical decision-making based on disease response to drugs, all at the single-cell level. The development of deep learning technology has provided a powerful tool for research in drug discovery based on single-cell techniques, which has evolved with the advent of large-scale public databases to predict drug responses and targets. In addition, traditional Chinese medicine (TCMs) research has also entered the era of single-cell technology. Single-cell omics technologies offer an alternative way in deciphering the mechanisms of TCMs in disease treatment, revealing drug targets, screening new drugs, and designing combinations of TCMs. This review aims to explore the application of single-cell omics technologies in drug screening and development comprehensively, highlighting how they accelerate the drug development process and facilitate personalized medicine by precisely identifying therapeutic targets, predicting drug responsiveness, deciphering mechanisms of action. It is also concluded that drug development process and therapeutic efficacy of drugs can be improved by combining single-cell omics and artificial intelligence techniques.

Interpretable Dynamic Directed Graph Convolutional Network for Multi-Relational Prediction of Missense Mutation and Drug Response.

Tumor heterogeneity presents a significant challenge in predicting drug responses, especially as missense mutations within the same gene can lead to varied outcomes such as drug resistance, enhanced sensitivity, or therapeutic ineffectiveness. These complex relationships highlight the need for advanced analytical approaches in oncology. Due to their powerful ability to handle heterogeneous data, graph convolutional networks (GCNs) represent a promising approach for predicting drug responses. However, simple bipartite graphs cannot accurately capture the complex relationships involved in missense mutation and drug response. Furthermore, Deep learning models for drug response are often considered "black boxes", and their interpretability remains a widely discussed issue. To address these challenges, we propose an Interpretable Dynamic Directed Graph Convolutional Network (IDDGCN) framework, which incorporates four key features: (1) the use of directed graphs to differentiate between sensitivity and resistance relationships, (2) the dynamic updating of node weights based on node-specific interactions, (3) the exploration of associations between different mutations within the same gene and drug response, and (4) the enhancement of interpretability models through the integration of a weighted mechanism that accounts for the biological significance, alongside a ground truth construction method to evaluate prediction transparency. The experimental results demonstrate that IDDGCN outperforms existing state-of-the-art models, exhibiting excellent predictive power. Both qualitative and quantitative evaluations of its interpretability further highlight its ability to explain predictions, offering a fresh perspective for precision oncology and targeted drug development.

Integrating molecular, biochemical, and immunohistochemical features as predictors of hepatocellular carcinoma drug response using machine-learning algorithms.

Liver cancer, particularly Hepatocellular carcinoma (HCC), remains a significant global health concern due to its high prevalence and heterogeneous nature. Despite the existence of approved drugs for HCC treatment, the scarcity of predictive biomarkers limits their effective utilization. Integrating diverse data types to revolutionize drug response prediction, ultimately enabling personalized HCC management. In this study, we developed multiple supervised machine learning models to predict treatment response. These models utilized classifiers such as logistic regression (LR), k-nearest neighbors (kNN), neural networks (NN), support vector machines (SVM), and random forests (RF) using a comprehensive set of molecular, biochemical, and immunohistochemical features as targets of three drugs: Pantoprazole, Cyanidin 3-glycoside (Cyan), and Hesperidin. A set of performance metrics for the complete and reduced models were reported including accuracy, precision, recall (sensitivity), specificity, and the Matthews Correlation Coefficient (MCC). Notably, (NN) achieved the best prediction accuracy where the combined model using molecular and biochemical features exhibited exceptional predictive power, achieving solid accuracy of 0.9693 ∓ 0.0105 and average area under the ROC curve (AUC) of 0.94 ∓ 0.06 coming from three cross-validation iterations. Also, found seven molecular features, seven biochemical features, and one immunohistochemistry feature as promising biomarkers of treatment response. This comprehensive method has the potential to significantly advance personalized HCC therapy by allowing for more precise drug response estimation and assisting in the identification of effective treatment strategies.

FOXN3-AS1: A Candidate Prognostic Marker and Epigenetic Target with Immunotherapeutic Implications in Acute Myeloid Leukemia.

We focused on the FOXN3 gene and selected its antisense transcripts (FOXN3-AS1) to investigate its potential involvement in acute myeloid leukemia (AML). Several integrated multi-omics datasets have expanded the horizons of the cancer landscape. With the emergence of new high-throughput technologies, a large number of non-coding RNAs have been confirmed to be involved in the pathogenesis of different types of hematological malignancies. We conducted experimental validation using quantitative polymerase chain reaction (qPCR) with bone marrow specimens from AML patients. Then, Kaplan-Meier (KM) and Receiver Operating Characteristic (ROC) curves were used to substantiate the prognostic association between FOXN3-AS1 and AML patients within the TCGA database. Correlation between FOXN3-AS1 expression and gene mutation, immune, and immune function using Spearman correlation analysis. To explore the physical and functional interaction between FOXN3-AS1 and the DNMT1 protein, we utilized the RPISeq web tool from Iowa State University. Subsequently, we performed qPCR experiments to test the effect of 5AzaC (DNMT1 inhibitor) on FOXN3-AS1 expression AML cell lines (THP1 and OCI-AML3). We leveraged the "OncoPredict" R package in conjunction with the Genomics of Drug Sensitivity (GDSC) database to predict drug response in AML patients expressing FOXN3-AS1. We observed a significant upregulation of FOXN3-AS1 expression in AML patients compared to healthy controls using clinical samples. The TCGA database revealed an association between high FOXN3-AS1 expression and adverse prognosis. In our subsequent analysis, genes with poor prognostic implications in AML patients were exclusively identified in the FOXN3-AS1 high-expression group, further corroborating this relationship. AML patients with higher FOXN3-AS1 expression levels may respond less optimally to immunotherapy than patients with lower levels. Besides, we computationally predicted the interaction of FOXN3- AS1 and DNMT1 protein and experimentally confirmed that DNMT1i (GSK-3484862) affects the expression level of FOXN3-AS1. We also found that the chemotherapy drugs (5-Fluorouralic, Cisplatin, Dactolisib, Sapitinib, Temozolomide, Ulixertinib, Vinorelbine, Ruxolitinib, Osimertinib and Cisplatin) showed favorable responses in AML patients with high FOXN3-AS1 expression levels. Our candidate approach identifies FOXN3-AS1 as a prognostic indicator of survival in AML with a potential immune-related role. The preliminary observations we made on FOXN3-AS1/DNMT1 crosstalk warrant more in-depth invested immunotherapeutic approaches in AML.

A Human Brain-Chip for Modeling Brain Pathologies and Screening Blood–Brain Barrier Crossing Therapeutic Strategies

Background/Objectives: The limited translatability of preclinical experimental findings to patients remains an obstacle for successful treatment of brain diseases. Relevant models to elucidate mechanisms behind brain pathogenesis, including cell-specific contributions and cell-cell interactions, and support successful targeting and prediction of drug responses in humans are urgently needed, given the species differences in brain and blood-brain barrier (BBB) functions. Human microphysiological systems (MPS), such as Organ-Chips, are emerging as a promising approach to address these challenges. Here, we examined and advanced a Brain-Chip that recapitulates aspects of the human cortical parenchyma and the BBB in one model. Methods: We utilized human primary astrocytes and pericytes, human induced pluripotent stem cell (hiPSC)-derived cortical neurons, and hiPSC-derived brain microvascular endothelial-like cells and included for the first time on-chip hiPSC-derived microglia. Results: Using Tumor necrosis factor alpha (TNFα) to emulate neuroinflammation, we demonstrate that our model recapitulates in vivo-relevant responses. Importantly, we show microglia-derived responses, highlighting the Brain-Chip’s sensitivity to capture cell-specific contributions in human disease-associated pathology. We then tested BBB crossing of human transferrin receptor antibodies and conjugated adeno-associated viruses. We demonstrate successful in vitro/in vivo correlation in identifying crossing differences, underscoring the model’s capacity as a screening platform for BBB crossing therapeutic strategies and ability to predict in vivo responses. Conclusions: These findings highlight the potential of the Brain-Chip as a reliable and time-efficient model to support therapeutic development and provide mechanistic insights into brain diseases, adding to the growing evidence supporting the value of MPS in translational research and drug discovery.

TransCDR: a deep learning model for enhancing the generalizability of drug activity prediction through transfer learning and multimodal data fusion

BackgroundAccurate and robust drug response prediction is of utmost importance in precision medicine. Although many models have been developed to utilize the representations of drugs and cancer cell lines for predicting cancer drug responses (CDR), their performances can be improved by addressing issues such as insufficient data modality, suboptimal fusion algorithms, and poor generalizability for novel drugs or cell lines.ResultsWe introduce TransCDR, which uses transfer learning to learn drug representations and fuses multi-modality features of drugs and cell lines by a self-attention mechanism, to predict the IC50 values or sensitive states of drugs on cell lines. We are the first to systematically evaluate the generalization of the CDR prediction model to novel (i.e., never-before-seen) compound scaffolds and cell line clusters. TransCDR shows better generalizability than 8 state-of-the-art models. TransCDR outperforms its 5 variants that train drug encoders (i.e., RNN and AttentiveFP) from scratch under various scenarios. The most critical contributors among multiple drug notations and omics profiles are Extended Connectivity Fingerprint and genetic mutation. Additionally, the attention-based fusion module further enhances the predictive performance of TransCDR. TransCDR, trained on the GDSC dataset, demonstrates strong predictive performance on the external testing set CCLE. It is also utilized to predict missing CDRs on GDSC. Moreover, we investigate the biological mechanisms underlying drug response by classifying 7675 patients from TCGA into drug-sensitive or drug-resistant groups, followed by a Gene Set Enrichment Analysis.ConclusionsTransCDR emerges as a potent tool with significant potential in drug response prediction.

Integrative analysis of multiple cell death model for precise prognosis and drug response prediction in gastric cancer.

Gastric cancer (GC) is a common upper gastrointestinal tumor. However, the evaluation of prognosis and treatment response in patients with gastric cancer remains a challenge. Programmed cell death (PCD) is one of the important terminal paths for the cells of metazoans, and is involved in a variety of biological events that include morphogenesis, maintenance of tissue homeostasis, and elimination of harmful cells. The objective of this project is to investigate the predictive significance of cell death pathways and create prognostic signatures associated to cell death, with the purpose of forecasting prognosis and providing guidance for the treatment of gastric cancer. Gene transcription profiles and corresponding clinical data of gastric cancer patients were collected from The Cancer Genome Atlas (TCGA-STAD, n = 448) and the Gene Expression Comprehensive Database (GSE84437, n = 483). Thirteen types of cell death-related genes, including apoptosis, necroptosis, pyroptosis, ferroptosis, autophagy, cuprotosis, parthanatos, entotic cell death, netotic cell death, lysosome-dependent cell death, alkaliptosis, oxeiptosis, and disulfidptosis, were analysed. Cell death-related genes associated with prognosis were identified in the TCGA-STAD training cohort using Lasso-Cox regression to generate a risk score. Patients were categorized into high and low-risk groups based on the median risk score for survival difference analysis. Cell death-related genes associated with prognosis were identified in the TCGA-STAD training cohort using Lasso-Cox regression to generate a risk score. Additionally, the response to immunotherapy in the high-risk and low-risk groups was calculated using the oncoPredict algorithm. Futhermore, the model genes were validated in the GEO validation set. A total of 324 differential programmed cell death (PCD)-related genes were identified, and 65 were selected through single-factor Cox analysis. Six PCD-related genes were ultimately identified by Lasso regression to construct a prognostic risk score model. The log-rank test revealed that patients in the high-risk group had inferior survival time compared with those in the low-risk group. The area under the ROC curve (AUC) for the training group at years 1, 3, and 5 were 0.684, 0.713, 0.743, respectively, while the AUC for the validation cohort at years 1, 3, and 5 were 0.695, 0.704, and 0.707, respectively. Unsupervised clustering identified potential subtypes included in the model, and a survival difference was also observed between the two subgroups. Multifactor Cox results, combined with clinical information, demonstrated that the prognostic risk score can serve as an independent prognostic factor, irrespective of other clinical features. By comprehensively analyzing multiple cell death patterns, we have established a novel model that accurately forecasts the clinical prognosis and drug sensitivity of gastric cancer. It was found that all 12 representative drugs may not be suitable for patients in high-risk groups.

DeepFusionCDR: Employing Multi-Omics Integration and Molecule-Specific Transformers for Enhanced Prediction of Cancer Drug Responses.

Deep learning approaches have demonstrated remarkable potential in predicting cancer drug responses (CDRs), using cell line and drug features. However, existing methods predominantly rely on single-omics data of cell lines, potentially overlooking the complex biological mechanisms governing cell line responses. This paper introduces DeepFusionCDR, a novel approach employing unsupervised contrastive learning to amalgamate multi-omics features, including mutation, transcriptome, methylome, and copy number variation data, from cell lines. Furthermore, we incorporate molecular SMILES-specific transformers to derive drug features from their chemical structures. The unified multi-omics and drug signatures are combined, and a multi-layer perceptron (MLP) is applied to predict IC50 values for cell line-drug pairs. Moreover, this MLP can discern whether a cell line is resistant or sensitive to a particular drug. We assessed DeepFusionCDR's performance on the GDSC dataset and juxtaposed it against cutting-edge methods, demonstrating its superior performance in regression and classification tasks. We also conducted ablation studies and case analyses to exhibit the effectiveness and versatility of our proposed approach. Our results underscore the potential of DeepFusionCDR to enhance CDR predictions by harnessing the power of multi-omics fusion and molecular-specific transformers. The prediction of DeepFusionCDR on TCGA patient data and case study highlight the practical application scenarios of DeepFusionCDR in real-world environments.

A disentangled generative model for improved drug response prediction in patients via sample synthesis

A disentangled generative model for improved drug response prediction in patients via sample synthesis

ENO2 in progression and treatment of colon adenocarcinoma: integrative bioinformatics analysis on non-apoptotic cell death

Colon adenocarcinoma (COAD) is one of the most common types of cancer. The interconnection between non-apoptotic cell death and COAD has not been adequately addressed. In our study, an integrative bioinformatics analysis was performed to explore non-apoptotic cell death-related biomarkers in COAD. ENO2 was determined as a potent biomarker for prognosis, drug response, immunity, and immunotherapy prediction. We used EdU and RT-qPCR assays to test our hypothesis and investigate how the ENO2 gene may influence or regulate cancer-related processes. ENO2 was expected to be a potential target in COAD.

Machine learning-based discovery of UPP1 as a key oncogene in tumorigenesis and immune escape in gliomas.

Gliomas are the most common and aggressive type of primary brain tumor, with a poor prognosis despite current treatment approaches. Understanding the molecular mechanisms underlying glioma development and progression is critical for improving therapies and patient outcomes. The current study comprehensively analyzed large-scale single-cell RNA sequencing and bulk RNA sequencing of glioma samples. By utilizing a series of advanced computational methods, this integrative approach identified the gene UPP1 (Uridine Phosphorylase 1) as a novel driver of glioma tumorigenesis and immune evasion. High levels of UPP1 were linked to poor survival rates in patients. Functional experiments demonstrated that UPP1 promotes tumor cell proliferation and invasion and suppresses anti-tumor immune responses. Moreover, UPP1 was found to be an effective predictor of mutation patterns, drug response, immunotherapy effectiveness, and immune characteristics. These findings highlight the power of combining diverse machine learning methods to identify valuable clinical markers involved in glioma pathogenesis. Identifying UPP1 as a tumor growth and immune escape driver may be a promising therapeutic target for this devastating disease.

Model ensembling as a tool to form interpretable multi-omic predictors of cancer pharmacosensitivity.

Stratification of patients diagnosed with cancer has become a major goal in personalized oncology. One important aspect is the accurate prediction of the response to various drugs. It is expected that the molecular characteristics of the cancer cells contain enough information to retrieve specific signatures, allowing for accurate predictions based solely on these multi-omic data. Ideally, these predictions should be explainable to clinicians, in order to be integrated in the patients care. We propose a machine-learning framework based on ensemble learning to integrate multi-omic data and predict sensitivity to an array of commonly used and experimental compounds, including chemotoxic compounds and targeted kinase inhibitors. We trained a set of classifiers on the different parts of our dataset to produce omic-specific signatures, then trained a random forest classifier on these signatures to predict drug responsiveness. We used the Cancer Cell Line Encyclopedia dataset, comprising multi-omic and drug sensitivity measurements for hundreds of cell lines, to build the predictive models, and validated the results using nested cross-validation. Our results show good performance for several compounds (Area under the Receiver-Operating Curve >79%) across the most frequent cancer types. Furthermore, the simplicity of our approach allows to examine which omic layers have a greater importance in the models and identify new putative markers of drug responsiveness. We propose several models based on small subsets of transcriptional markers with the potential to become useful tools in personalized oncology, paving the way for clinicians to use the molecular characteristics of the tumors to predict sensitivity to therapeutic compounds.

BANDRP: a bilinear attention network for anti-cancer drug response prediction based on fingerprint and multi-omics.

Predicting anti-cancer drug response can help with personalized cancer treatment and is an important topic in modern oncology research. Although some methods have been used for anti-cancer drug response prediction, how to effectively integrate various features related to cancer cell lines, drugs, and their known responses is still affected by the redundant information of input features and the complex interactions between features. In this study, we propose a bilinear attention model, named BANDRP, based on multiple omics data of cancer cell lines and multiple molecular fingerprints of drugs to predict potential anti-cancer drug responses. Compared with existing models, BANDRP uses gene expression data to calculate pathway enrichment scores to enrich the features of cancer cell lines and can automatically learn the interactive information of cancer cell lines and drugs through bilinear attention networks. Benchmarking and independent tests demonstrate that BANDRP surpasses baseline models and exhibits robust generalization performance. Ablation experiments affirm the optimality of the current model architecture and feature selection scheme for our prediction task. Furthermore, analytical experiments and case studies on unknown anti-cancer drug response predictions underscore BANDRP's potential as a potent and reliable framework for predicting anti-cancer drug response.

Multi-output prediction of dose–response curves enables drug repositioning and biomarker discovery

Drug response prediction is hampered by uncertainty in the measures of response and selection of doses. In this study, we propose a probabilistic multi-output model to simultaneously predict all dose–responses and uncover their biomarkers. By describing the relationship between genomic features and chemical properties to every response at every dose, our multi-output Gaussian Process (MOGP) models enable assessment of drug efficacy using any dose–response metric. This approach was tested across two drug screening studies and ten cancer types. Kullback-leibler divergence measured the importance of each feature and identified EZH2 gene as a novel biomarker of BRAF inhibitor response. We demonstrate the effectiveness of our MOGP models in accurately predicting dose–responses in different cancer types and when there is a limited number of drug screening experiments for training. Our findings highlight the potential of MOGP models in enhancing drug development pipelines by reducing data requirements and improving precision in dose–response predictions.