Clinical Drug Response Research Articles

CD27 as a Diagnostic Biomarker and Its Role in Immune Heterogeneity and Predicting Clinical Drug Responses in Hashimoto’s Thyroiditis

CD27 as a Diagnostic Biomarker and Its Role in Immune Heterogeneity and Predicting Clinical Drug Responses in Hashimoto’s Thyroiditis

TMOD-33. DOES PRE-CLINICAL DRUG ACTIVITY PREDICT PHASE II CLINICAL EFFICACY IN GLIOBLASTOMA? A DETAILED ANALYSIS

Abstract Despite decades of research, there are fewer than six FDA-approved therapies for glioblastoma (GBM). The drug approval process relies on studies establishing pre-clinical activity, most commonly using murine models, followed by in human studies establishing safety and then efficacy. Although murine models for GBM have allowed for insights into tumor biology, their utility in assessing the efficacy of novel anti-cancer therapies remains highly debated. Here, we sought to establish whether there was any relationship between drug activity in preclinical models and phase 2 clinical trial outcomes. Using ClinicalTrials.gov (NIH), all phase 2 trials in glioblastoma up to December 2023 were queried and downloaded. Trials that were never started, had not yet started, were withdrawn, not completed, or contained indications beyond glioblastoma were excluded. Studies examining anti-tumor effects of an investigational agent with or without bevacizumab, temozolomide, or radiation therapy in glioblastoma were included, and studies examining alternative dosing strategies, various medical devices, or drug combinations, were excluded. The corresponding preclinical studies were identified via literature review for each clinical trial. Data including the progression free and overall survival benefit, mouse models used, and sample sizes were extracted from each study. 659 trials between 1994 and 2023 were included for analysis. Of these, 272 (41%) were for monotherapies, examining 90 unique small molecules. Preclinical endpoints included murine survival and tumor weight. In comparing mouse survival and tumor size to progression free and overall survival in humans, our preliminary analysis did not reveal any clear correlation. The correlation between preclinical and clinical drug response in GBM remains uncertain. Future work will examine whether mouse genotype, immune status, or GBM cell line used affected the degree of concordance with human clinical trial data. Our analysis suggests that further investigation into preclinical models of GBM, and standardized reporting of preclinical models, are needed.

Machine learning-based biomarker screening for acute myeloid leukemia prognosis and therapy from diverse cell-death patterns

Acute myeloid leukemia (AML) exhibits pronounced heterogeneity and chemotherapy resistance. Aberrant programmed cell death (PCD) implicated in AML pathogenesis suggests PCD-related signatures could serve as biomarkers to predict clinical outcomes and drug response. We utilized 13 PCD pathways, including apoptosis, pyroptosis, ferroptosis, autophagy, necroptosis, cuproptosis, parthanatos, entotic cell death, netotic cell death, lysosome-dependent cell death, alkaliptosis, oxeiptosis, and disulfidptosis to develop predictive models based on 73 machine learning combinations from 10 algorithms. Bulk RNA-sequencing, single-cell RNA-sequencing transcriptomic data, and matched clinicopathological information were obtained from the TCGA-AML, Tyner, and GSE37642-GPL96 cohorts. These datasets were leveraged to construct and validate the models. Additionally, in vitro experiments were conducted to substantiate the bioinformatics findings. The machine learning approach established a 6-gene pan-programmed cell death-related genes index (PPCDI) signature. Validation in two external cohorts showed high PPCDI associated with worse prognosis in AML patients. Incorporating PPCDI with clinical variables, we constructed several robust prognostic nomograms that accurately predicted prognosis of AML patients. Multi-omics analysis integrating bulk and single-cell transcriptomics revealed correlations between PPCDI and immunological features, delineating the immune microenvironment landscape in AML. Patients with high PPCDI exhibited resistance to conventional chemotherapy like doxorubicin but retained sensitivity to dasatinib and methotrexate (FDA-approved drugs for other leukemias), suggesting the potential of PPCDI to guide personalized therapy selection in AML. In summary, we developed a novel PPCDI model through comprehensive analysis of diverse programmed cell death pathways. This PPCDI signature demonstrates great potential in predicting clinical prognosis and drug sensitivity phenotypes in AML patients.

Cross-Domain Feature Disentanglement for Interpretable Modeling of Tumor Microenvironment Impact on Drug Response.

High-throughput screening technology has enabled the generation of large-scale drug responses across hundreds of cancer cell lines. There remains a significant gap between in vitro cell lines and actual tumors in vivo in terms of their response to drug treatments yet. This is because tumors consist of a complex cellular composition and histopathology structure, known as the tumor microenvironment (TME), which greatly impacts the drug cytotoxicity against tumor cells. To date, no study has focused on modeling the impact of the TME on clinical drug response. In this study, we postulated that the intricate complexity of an actual tumor can be conceptually simplified into two separable components: cancerous cells and the tumor microenvironment. This assumption allowed us to model the influence of these two constituent parts on drug response through feature disentanglement. We employed a domain adaptation network to decouple and extract features from tumor transcriptional profiles. Specifically, two denoising autoencoders were separately used to extract features from cell lines (source domain) and tumors (target domain) for partial domain alignment and feature decoupling. The private encoder was enforced to extract information only about the TME. Moreover, to ensure generalizability to novel drugs, we employed a graph attention network to learn the latent representation of drugs, enabling us to linearly model the drug perturbation on cellular state in latent space. We validated our model on a benchmark dataset and demonstrated its superior performance in predicting clinical drug response and dissecting the influence of the TME on drug efficacy.

Evaluation of tumor response to immune checkpoint inhibitors by a 3D immunotumoroid model.

Only 20 percent of renal and bladder cancer patients will show a significant response to immune checkpoint inhibitor (ICI) therapy, and no test currently available accurately predicts ICI response. We developed an "immunotumoroid" cell model system that recapitulates the tumor, its microenvironment, and necessary immune system components in patient-derived spheroids to enable ex vivo assessment of tumor response to ICI therapy. Immunotumoroids were developed from surgically resected renal cell carcinomas and bladder carcinomas selected for high tumor-infiltrating lymphocytes (TILs) and survived more than a month without media exchange. Immunohistochemistry was used to detect immune and non-immune cells in cryopreserved source tumors and the resulting immunotumoroids. Immunotumoroid response to ICIs (nivolumab, pembrolizumab, and durvalumab) and chemotherapy (cisplatin, gemcitabine, and paclitaxel) was monitored in real-time with Cytotox Red staining in an Incucyte device, and the immunotumoroid response was compared to retrospective clinical drug responses. Six of the 13 cases tested grew viable immunotumoroid models, with failed cases attributed to extensive tumor tissue necrosis or excess lymphocytes preventing spheroid formation. One successfully cultured case was excluded from the study due to low TIL infiltration (<5%) in the primary tumor sample. The five remaining models contained immune cells (CD4+ and CD8+ T cells, and macrophages), non-immune cells (fibroblasts), and tumor cells. Chemotherapy and ICI drugs were tested in immunotumoroids from 5 cases and compared to clinical outcomes where data was available. Four/five models showed cell killing in response to chemotherapy and two/five showed sensitivity to ICI. In three cases, the immunotumoroid model accurately predicted the patient's clinical response or non-response to ICIs or chemotherapy. Our immunotumoroid model replicated the multicellular nature of the tumor microenvironment sufficiently for preclinical ICI screening. This model could enable valuable insights into the complex interactions between cancer cells, the immune system, and the microenvironment. This is a feasibility study on a small number of cases, and additional studies with larger case numbers are required including correlation with clinical response.

Differential whole-genome doubling based signatures for improvement on clinical outcomes and drug response in patients with breast cancer

Whole genome doublings (WGD), a hallmark of human cancer, is pervasive in breast cancer patients. However, the molecular mechanism of the complete impact of WGD on survival and treatment response in breast cancer remains unclear. To address this, we performed a comprehensive and systematic analysis of WGD, aiming to identify distinct genetic alterations linked to WGD and highlight its improvement on clinical outcomes and treatment response for breast cancer. A linear regression model along with weighted gene co-expression network analysis (WGCNA) was applied on The Cancer Genome Atlas (TCGA) dataset to identify critical genes related to WGD. Further Cox regression models with random selection were used to optimize the most useful prognostic markers in the TCGA dataset. The clinical implication of the risk model was further assessed through prognostic impact evaluation, tumor stratification, functional analysis, genomic feature difference analysis, drug response analysis, and multiple independent datasets for validation. Our findings revealed a high aneuploidy burden, chromosomal instability (CIN), copy number variation (CNV), and mutation burden in breast tumors exhibiting WGD events. Moreover, 247 key genes associated with WGD were identified from the distinct genomic patterns in the TCGA dataset. A risk model consisting of 22 genes was optimized from the key genes. High-risk breast cancer patients were more prone to WGD and exhibited greater genomic diversity compared to low-risk patients. Some oncogenic signaling pathways were enriched in the high-risk group, while primary immune deficiency pathways were enriched in the low-risk group. We also identified a risk gene, ANLN (anillin), which displayed a strong positive correlation with two crucial WGD genes, KIF18A and CCNE2. Tumors with high expression of ANLN were more prone to WGD events and displayed worse clinical survival outcomes. Furthermore, the expression levels of these risk genes were significantly associated with the sensitivities of BRCA cell lines to multiple drugs, providing valuable insights for targeted therapies. These findings will be helpful for further improvement on clinical outcomes and contribution to drug development in breast cancer.

Abstract 3500: Refining diagnosis of renal cell carcinoma subtypes through single-cell resolution transcriptomic signatures

Abstract A remaining challenge in the field of renal cell carcinoma (RCC) is diagnosis of rare or mixed subtypes, which hampers the delivery of the most effective healthcare to patients. While a new classification scheme, integrating molecular characteristics of tumors, is emerging, it requires a deep understanding of the tumor heterogeneity and etiopathogenesis of each subtype. We have generated and analyzed single-cell transcriptome profiles of 38 human RCCs and normal kidneys 16 RCCs with diverse histological subtypes and established gene signatures specific to each cell type. We have applied these gene signatures to bulk RNA-sequencing data of 1000 RCC tumors with different histological subtypes and normal kidneys from The Cancer Genome Atlas (TCGA) to characterize different cancer types and cellular origins. Unsupervised clustering analysis of TCGA data shows diagnostic discrepancies for 9.3% cases and subclassifies each major subtype (clear cell, papillary and chromophobe) into clusters with differential clinical outcomes.[KM1]. The pathological marker expression in misdiagnosed tumors is consistent with clustering-based classification, not their original diagnosis. Notably, using unique and subtype-specific gene signature patterns, we identified clear cell papillary renal cell tumor (ccpRCT), which is a newly classified benign entity and the main tumor type representing the mis-diagnosed tumors. These observations have further been confirmed by reviewing histological images of the tumors by an experienced genitourinary pathologist. Interestingly, our gene signatures reveal that ccpRCT does not originate from proximal tubule cells unlike clear cell and papillary RCCs. Taken together, our approach suggests a potential new classification method connected to clinical behaviors and drug responses based on overall transcriptomic differences. Citation Format: Minjun Kim, Ariel Madrigal, Zohreh Mehrjoo, Kate I. Glennon, Madeleine Arseneault, Morag Park, Fadi Brimo, Simon Tanguay, Hamed S. Najafabadi, Yasser Riazalhosseini. Refining diagnosis of renal cell carcinoma subtypes through single-cell resolution transcriptomic signatures [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 3500.

Abstract 907: ScreenDL: A transfer learning framework integrating tumor omics and functional drug screening for personalized clinical drug response prediction

Abstract Precision oncology hinges on accurate prediction of patient-specific treatment response from tumor molecular inputs. While deep learning (DL) models achieve state-of-the-art response prediction in cell lines, existing methods do not readily translate to the clinic where training data is limited. Here, we have developed and validated ScreenDL, a novel DL framework designed explicitly for use in clinical precision oncology applications. The underlying architecture of ScreenDL consists of two fully connected branches dedicated to learning rich embeddings of a drug’s chemical structure and a tumor’s transcriptomic profile respectively. These tumor omic and drug molecular embeddings are fed into a shared response prediction subnetwork which outputs predicted ln(IC50) values. The ScreenDL training schema incorporates three phases, each designed to improve clinical response prediction: (1) an initial pretraining phase leverages large-scale cell line pharmacogenomic datasets to extract patterns/relationships linking tumor omics and drug response; (2) subsequent transfer learning integrates pharmacogenomic data from patient-derived models of cancer, adapting the pretrained model to a more patient-relevant context; (3) patient-level omics and drug screening data (e.g., from matched patient-derived organoid (PDO) models) is integrated through patient-specific fine-tuning, generating a patient-specific response prediction model. Importantly, biomaterial from surgical tumor resection is often available for organoid establishment and patient-specific drug screening in practical clinical scenarios. To assess the utility of our ScreenDL framework for clinical response prediction, we applied ScreenDL to predict treatment response in 50 advanced/metastatic triple negative breast cancer (TNBC) patient-derived xenograft models (PDX). We leveraged tumor omics profiles for each PDX and drug screening data from matched PDX-derived organoids (PDxOs), mirroring the combination of patient-level tumor omic characterization and functional drug screening in matched PDOs, a protocol currently being piloted in Functional Precision Oncology trials at the University of Utah Huntsman Cancer Institute. After cell line pretraining, ScreenDL achieves a modest median Pearson correlation per drug of 0.15 relative to 0.03 for the leading compared model. However, transfer learning and subsequent PDX-specific fine-tuning significantly improve performance, producing median Pearson correlations per drug of 0.39 and 0.51, respectively. These results demonstrate the ability of our ScreenDL framework to dramatically improve response prediction in clinically relevant domains, bringing DL-based precision treatment selection closer to clinical application. Citation Format: Casey Sederman, Tonya Di Sera, Yi Qiao, Xiaomeng Huang, Bryan E. Welm, Alana L. Welm, Gabor Marth. ScreenDL: A transfer learning framework integrating tumor omics and functional drug screening for personalized clinical drug response prediction [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 907.

Multilayered insights: a machine learning approach for personalized prognostic assessment in hepatocellular carcinoma.

Hepatocellular carcinoma (HCC) is a complex malignancy, and precise prognosis assessment is vital for personalized treatment decisions. This study aimed to develop a multi-level prognostic risk model for HCC, offering individualized prognosis assessment and treatment guidance. By utilizing data from The Cancer Genome Atlas (TCGA) and the Surveillance, Epidemiology, and End Results (SEER) database, we performed differential gene expression analysis to identify genes associated with survival in HCC patients. The HCC Differential Gene Prognostic Model (HCC-DGPM) was developed through multivariate Cox regression. Clinical indicators were incorporated into the HCC-DGPM using Cox regression, leading to the creation of the HCC Multilevel Prognostic Model (HCC-MLPM). Immune function was evaluated using single-sample Gene Set Enrichment Analysis (ssGSEA), and immune cell infiltration was assessed. Patient responsiveness to immunotherapy was evaluated using the Immunophenoscore (IPS). Clinical drug responsiveness was investigated using drug-related information from the TCGA database. Cox regression, Kaplan-Meier analysis, and trend association tests were conducted. Seven differentially expressed genes from the TCGA database were used to construct the HCC-DGPM. Additionally, four clinical indicators associated with survival were identified from the SEER database for model adjustment. The adjusted HCC-MLPM showed significantly improved discriminative capacity (AUC=0.819 vs. 0.724). External validation involving 153 HCC patients from the International Cancer Genome Consortium (ICGC) database verified the performance of the HCC-MLPM (AUC=0.776). Significantly, the HCC-MLPM exhibited predictive capacity for patient response to immunotherapy and clinical drug efficacy (P < 0.05). This study offers comprehensive insights into HCC prognosis and develops predictive models to enhance patient outcomes. The evaluation of immune function, immune cell infiltration, and clinical drug responsiveness enhances our comprehension and management of HCC.

Abstract B060: Enhancing models of pancreatic cancer: Integrating patient-derived models with tissue engineering to investigate clinical responses to therapy

Abstract Cancer modeling systems can be advantageous in enabling personalized medicine through rapid testing of therapeutic responsiveness. However, modeling systems often fail in these efforts due to 1) inadequately replicating unique patient genetics, 2) inadequately modeling the patient tumor microenvironment (TME), and 3) lack of internal validation references when isolated from clinical drug response data. We have developed a patient-derived models of cancer (PDMC) program for pancreatic cancer that addresses these challenges. By integrating patient cancer modeling and tissue engineering efforts with clinical trial protocols that perform serial tissue biopsies and analyses during treatment, we can make direct assessments of how different modeling systems replicate actual patient outcomes. Clinical trial protocols include tissue biopsies pre- and post- treatment using targeted therapeutics including MEK and PARP inhibitors. The modeling program, meanwhile, includes collecting metadata from pre- and post- treatment biopsies, the banking of patient tissues, cell line and xenograft generation from each biopsy, multiomic comparisons of models to patient tissue, comparisons of therapeutic responsiveness in models, and ultimately tissue re-engineering through bioprinting of tumor cells along with TME components. Whole exome sequencing is used to confirm that genetic alterations are preserved in each model, and we have established models representing a range of human PDAC genomic profiles. This includes common KRAS and TP53 mutations, as well as rare BRCA mutations and KRAS-wild-type tumors with BRAF mutations. Many cancer-associated fibroblast (CAF) lines have also been established through this program. Including patient derived cell lines, CAD lines, endothelial cells, and PBMCs into 3D bioprinted tumors allows us to better model the TME. Assays performed upon bioprinted tissues can then be directly compared to the same assays performed on the patient tumor to assess similarities as well as further obstacles. Ongoing efforts include building a resource of precise and validated PDMCs for diverse pancreatic cancers, and establishing a path forward for improved cancer modeling for the actualization of personalized medicine in PDAC treatment. Citation Format: Alexander C. Smith, John Muschler, Mike Munks, Rosalie C. Sears, Jonathan Brody. Enhancing models of pancreatic cancer: Integrating patient-derived models with tissue engineering to investigate clinical responses to therapy [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Pancreatic Cancer; 2023 Sep 27-30; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(2 Suppl):Abstract nr B060.

Predicting Clinical Anticancer Drug Response of Patients by using Domain Alignment and Prototypical Learning.

Anticancer drug response prediction is crucial in developing personalized treatment plans for cancer patients. However, High-quality patient anticancer drug response data are scarce and cell line data and patient data have different distributions, models trained solely on cell line data perform poorly. Some existing methods predict anticancer drug response by transferring knowledge from the cell line domain to the patient domain using transfer learning. However, the robustness of these classifiers is affected by anomalies in the cell line data, and they do not utilize the knowledge in the unlabeled target domain data. To this end, we proposed a model called DAPL to predict patient responses to anticancer drugs. The model extracts domain-invariant features from cell lines and patients by constructing multiple VAEs and extracts drug features using GNNs. These features are then combined for prototypical learning to train a classifier, resulting in better predictions of patient anticancer drug response. We used the cell line datasets CCLE and GDSC as source domains and the patient datasets TCGA and PDTC as target domains and conducted experiments. The results indicate that DAPL shows excellent performance in predicting patient anticancer drug response compared to other state-of-the-art methods.

Identification of hub genes to determine drug-disease correlation in breast carcinomas.

The exact molecular mechanism underlying the heterogeneous drug response against breast carcinoma remains to be fully understood. It is urgently required to identify key genes that are intricately associated with varied clinical response of standard anti-cancer drugs, clinically used to treat breast cancer patients. In the present study, the utility of transcriptomic data of breast cancer patients in discerning the clinical drug response using machine learning-based approaches were evaluated. Here, a computational framework has been developed which can be used to identify key genes that can be linked with clinical drug response and progression of cancer, offering an immense opportunity to predict potential prognostic biomarkers and therapeutic targets. The framework concerned utilizes DeSeq2, Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), Cytoscape, and machine learning techniques to find these crucial genes. Total RNA extraction and qRT-PCR were performed to quantify relative expression of few hub genes selected from the networks. In our study, we have experimentally checked the expression of few key hub genes like APOA2, DLX5, APOC3, CAMK2B, and PAK6 that were predicted to play an immense role in breast cancer tumorigenesis and progression in response to anti-cancer drug Paclitaxel. However, further experimental validations will be required to get mechanistic insights of these genes in regulating the drug response and cancer progression which will likely to play pivotal role in cancer treatment and precision oncology.

Tamoxifen induces ferroptosis in MCF-7 organoid.

Breast cancer is the most common female malignant tumor type globally. The occurrence and development of breast cancer involve ferroptosis, which is closely related to its treatment. The development of breast cancer organoids facilitates the analysis of breast cancer molecular background and tumor biological behavior, including clinical pathological characteristics, drug response, or drug resistance relationship, and promotes the advancement of precision treatment for breast cancer. The three-dimensional (3D) cell culture of breast cancer MCF-7 organoid is more similar to the in vivo environment and thus obtains more realistic results than 2D cell culture. Our study examined the new mechanism of tamoxifen in treating breast cancer through breast cancer MCF-7 organoids. We used 3D cells to culture breast cancer MCF-7 organoid, as well as tamoxifen-treated MCF-7 and tamoxifen-resistant MCF-7 (MCF-7 TAMR) cells. We used transcriptome sequencing. We detected GPX4 and SLC7A11 protein levels using Western blotting and the content of ATP, glutathione, and ferrous ions using the Cell Counting Lite 3D Kit. We assessed cell viability using the Cell Counting Kit-8 (CCK-8) assay. Tamoxifen significantly inhibited the growth of MCF-7 organoids and significantly induced ferroptosis in MCF-7 organoids. The ferroptosis inhibitor reversed the significant tamoxifen-induced MCF-7 organoid inhibition activity. Moreover, the ferroptosis activator enhanced the tamoxifen-induced MCF-7 TAMR cell activity inhibition. Our study revealed that ferroptosis plays an important role in tamoxifen-induced MCF-7 organoid cell death and provides a new research idea for precise treatment of breast cancer through an organoid model.

Abstract C176: Cancer-specific AI identifies multi-modal biomarkers of therapeutic response for 1,951 drugs including TNG348, a highly selective USP1 inhibitor

Abstract Understanding drug mechanism-of-action and improving clinical response rates are critical challenges to cancer therapeutic development. We designed a deep learning framework that learns molecular mechanisms of drug sensitivity and resistance from preclinical and/or clinical drug response data. The trained model can then generate biomarker strategies, suggest companion diagnostics, and propose rational drug-drug combinations that increase or restore drug sensitivity. The key innovation is the model’s architecture; the framework learns an embedding that reflects an interpretable relationship between a cellular process and drug response by emulating a tumor cell’s proteomic architecture. We applied the framework to learn genetic determinants of response for 1,951 drugs and 17,386 CRISPR gene knockouts screened in human cancer cell lines. The learned embeddings reflected known and novel mechanisms. RAF inhibitors, MEK inhibitors, and BRAF CRISPR knockout colocalized (P=1.7×10-11) due to their shared dependency on hyperactivation of RAF signaling. CDK4/6 inhibitor efficacy was strongly associated with a set of physically interacting proteins that control G1-to-S transition (P=5.6×10-12). Validating the clinical applicability of the finding, the trained model stratified patient response in a retrospective analysis of 70 ER+ metastatic breast cancer patients treated with the CDK4/6 inhibitor palbociclib (P=0.05, log-rank predicted responsive vs. non-responsive). As a case study, we examined mechanisms of sensitivity and resistance to TNG348 – a potent and highly selective USP1 inhibitor – in non-small cell lung cancer (NSCLC). By integrating genetics, gene expression, and CRISPR screens, the model identified seven genes whose expression explained &amp;gt;50% of the variation in response to TNG348 across 32 held-out NSCLC cell lines (Spearman R=0.74; P=8.7×10-7). The findings extend TNG348’s known relationship with homologous repair to additional protein complexes involved in cell cycle and DNA repair. The model leveraged the new associations to generate a six-biomarker inclusion-exclusion logic that captures both a large population (31.4% of TCGA lung adenocarcinoma patients meet inclusion criteria) and substantially enriches for strong response amongst preclinical models (OR=14.8, P=3.0×10-3). Finally, we asked the model to design rational combinations to overcome acquired resistance mechanisms. Proposed combinations included clinically successful combinations such as RAF+MEK inhibitors and were enriched for synergistic matches in a cell line screen of 986 drug pairings (Spearman rho P=6.6×10-4). These results highlight the promise of explainable AI to learn complex cellular mechanisms-of-action and generate non-obvious biomarker hypothesis and rational drug combination strategies for clinical development. Citation Format: Adam Yaari, Lee McDaniel, Antoine Simoneau, Samuel Meier, Oliver Priebe, Eduardo Farias, Alan Huang, Jannik Andersen, Yi Yu, Maxwell Sherman. Cancer-specific AI identifies multi-modal biomarkers of therapeutic response for 1,951 drugs including TNG348, a highly selective USP1 inhibitor [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2023 Oct 11-15; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2023;22(12 Suppl):Abstract nr C176.

Integrative Multi-Omic Analysis for Prognosis Stratification in Acute Myeloid Leukemia

Acute myeloid leukemia (AML) is a heterogeneous hematopoietic malignancy with a dismal prognosis. European LeukemiaNet (ELN) is crucial for tailoring AML treatment individually. Various AML models correlated survival and clinical drug response with immune cell differentiation state by deconvoluting transcriptomics. However, few comprehensive molecular subtyping models integrate multi-omic profiles for prognostic and drug response prediction. Thus, this study aimed to integrate DNA methylation, genomics, transcriptomics and ex vivo drug sensitivity screening of AML patients and explore their roles in facilitating AML molecular stratification and redefining prognosis. Firstly，data on 90 patients with AML(non-promyelocytic leukemia) were retrieved from The Cancer Genome Atlas (TCGA) for consensus clustering by 10 advanced consensus clustering algorithms. The optimal three distinct subtypes, UAMOCS1, UAMOCS2, and UAMCOS3, were identified to achieve the best clustering effect. Multi-omics were orchestrated together for UAMOCS shaping. The UAMOCS can significantly distinguish overall survival (median 18 vs. 27 vs. 45 months, P= 0.0048) among the three instinct subtypes with the integrated contribution of transcriptomic profiles and DNA methylation and mutation data. Also, the activity of AML-specific transcription factors and chromatin remodeling makers showed a diverse pattern among the three subtypes, epigenetically contributing to UAMOCS shaping. Next, the nearest template prediction, a model-free method, was applied to further validate this predictive model. Similar subtyping under this model is generated robustly by imitating 50 subtype-specific genes and capable of identifying survival in both the “ihCAMs-AML” (Figure 1) and GSE37642. The main validation cohort was a real-world de novo AML cohort (acute promyelocytic leukemia excluded), “ihCAMs-AML”（N=98）, aged 14-70 years, with a median follow-up of 28 (1-55) months. The samples were simultaneously available for methylome (N=31, 850K Methylation Chip), DNA mutation (N=96, targeted deep sequencing), transcriptome (N=98, bulk RNA-seq), and ex vivo drug screening (N=66, 17 drugs under 100% peak plasma concentration) profiling analysis. Clinical relevance showed that traditional cytogenetic stratification was strongly correlated with our system. UAMOCS1, acting as an “immune activator”, is characterized by enriching of AML myelodysplasia-related (AML-MR) gene mutations (especially RUNX1 mutation, P = 0.00839), unstable chromosome (deletions on 7q), increased abundance of immune cells with immune escaping marker and poor risk. UAMOCS2 was defined as an intermediate immune burden and a “monocytic-like” phenotype with intermediate survival, corresponding to its inclining result toward M4/M5. UAMOCS3 was an “immune desert” group supposed to be the most favorable subset among the three groups. The ex vivo drug screening result implied that UAMOCS also provide values in distinguishing drug response. UAMOCS1 hardly benefited from commonly used chemotherapeutic drugs compared with the other two distinct subtypes. Pharmacogenomics database GDSC (the Genomics of Drug Sensitivity in Cancer) among AML cell lines was used to evaluate potential drug targets for overcoming drug resistance. Receptor tyrosine kinase signaling inhibitors (SB505124 and EphB4) might be sensitive to UAMOCS1. Taken together, UAMOCS model was generated that can classify AML patients independently and help develop precise chemotherapeutic strategy for patients with AML in the future.

Emden: A novel method integrating graph and transformer representations for predicting the effect of mutations on clinical drug response

Precision medicine based on personalized genomics provides promising strategies to enhance the efficacy of molecular-targeted therapies. However, the clinical effectiveness of drugs has been severely limited due to genetic variations that lead to drug resistance. Predicting the impact of missense mutations on clinical drug response is an essential way to reduce the cost of clinical trials and understand genetic diseases. Here, we present Emden, a novel method integrating graph and transformer representations that predicts the effect of missense mutations on drug response through binary classification with interpretability. Emden utilized protein sequences-based features and drug structures as inputs for rapid prediction, employing competitive representation learning and demonstrating strong generalization capabilities and robustness. Our study showed promising potential for clinical drug guidance and deep insight into computer-assisted precision medicine. Emden is freely available as a web server at https://www.psymukb.net/Emden.

Immune-related gene-based model predicts the survival of colorectal carcinoma and reflected various biological statuses.

Bakcground: Prognosis of colorectal cancer (CRC) varies due to complex genetic-microenviromental interactions, and multiple gene-based prognostic models have been highlighted. Material and Method: In this work, the immune-related genes' expression-based model was developed and the scores of each sample were calculated. The correlation between the model and clinical information, immune infiltration, drug response and biological pathways were analyzed. Results: The high-score samples have a significantly longer survival (overall survival and progression-free survival) period than those with a low score, which was validated across seven datasets containing 1,325 samples (GSE17536 (N = 115), GSE17537 (N = 55), GSE33113 (N = 90), GSE37892 (N = 130), GSE38832 (N = 74), GSE39582 (N = 481), and TCGA (N = 380)). The score is significantly associated with clinical indicators, including age and stage, and further associated with PD-1/PD-L1 gene expression. Furthermore, high-score samples have significantly higher APC and a lower MUC5B mutation rate. The high-score samples show more immune infiltration (including CD4+ and CD8+ T cells, M1/M2 macrophages, and NK cells). Enriched pathway analyses showed that cancer-related pathways, including immune-related pathways, were significantly activated in high-score samples and that some drugs have significantly lower IC50 values than those with low score. Conclusion: The model developed based on immune-related genes is robust and reflected various statuses of CRC and may be a potential clinical indicator.

712 Correlations between in-vitro and clinical drug response to CFTR modulators

712 Correlations between in-vitro and clinical drug response to CFTR modulators

Pan-Cancer Multi-Omics Analysis of Minichromosome Maintenance Proteins (MCMs) Expression in Human Cancers.

Epigenetic modifications, such as transcription, DNA repair, and replication significantly influence tumour development. Aberrant gene expression and modifications can have a crucial impact on the initiation and progression of tumours. The minichromosome maintenance (MCM) protein family, which is responsible for DNA synthesis, plays a crucial role in tumorigenesis and chemotherapy resistance by regulating the cell cycle and DNA replication stress. Recent studies have shown that dysregulation of the MCMs can lead to these negative outcomes. This study aimed to examine the role of the MCM proteins in DNA synthesis in 33 types of cancers. Various public databases were used to examine the expression, methylation regulation, mutations, and functions of eight MCM proteins (MCM2-9) in pan-cancer. The study investigated the correlation between abnormal MCM expression and clinical outcomes, including prognosis and drug response. The microRNA-mRNA network upstream of the MCM genes and the downstream signalling pathways were extensively investigated to determine the molecular mechanisms that drive tumour development. The study found that the MCM gene expressions differed depending on the type of cancer; high MCM gene expression was linked to poor overall survival in most cancers. Additionally, MCM gene expression was associated with various immunological features and drug sensitivity. These findings offer important insights for the development of targeted cancer therapies. Altogether, this study reveals that the MCM genes are differentially expressed across various cancers and are associated with clinical prognoses. These genes may influence the occurrence and development of tumours through several pathways, including the PI3K-AKT, PAS/MAPK and TSC/mTOR signalling pathways and immune-related pathways.

Colorectal cancer tissue-originated spheroids reveal tumor intrinsic signaling pathways and mimic patient clinical chemotherapeutic response as a rapid and valid model

Locally advanced colorectal cancer requires preoperative chemotherapy to reduce local recurrence and metastasis rates, but it remains difficult to predict the tumor will be sensitive to which treatments. The patient-derived organoids (PDOs) are considered an effective platform for predicting tumor drug responses in precision oncology. However, it has the limitation of being time-consuming in practical applications, especially in neoadjuvant treatment. Here we used cancer tissue-originated spheroids (CTOS) method to establish organoids from a heterogeneous population of colorectal cancer specimens, and evaluated the capacity of CTOS to predict clinical drug responses. By analyzing the relationship of the activities of drug-treated CTOS, drug targets and target-related pathways, tumor intrinsic effective-target-related pathways can be identified. These pathways were highly matched to the abnormal pathways indicated by whole-exome sequencing. Based on this, we used half effective concentration gradients to classify CTOS as sensitive or resistant to chemotherapy regimens within a week, for predicting neoadjuvant treatment outcomes for colorectal cancer patients. The drug sensitivity test results are highly matched to the clinical responses to treatment in individual patients. Thus, our data suggested that CTOS models can be effectively screened ex vivo to identify pathways sensitive to chemotherapies. These data also supported organoid research for personalized clinical medication guidance immediately after diagnosis in patients with advanced colorectal cancer.