Drug Sensitivity Score Research Articles

Development of biomarker signatures associated with anoikis to predict prognosis in patients with esophageal cancer: An observational study.

Anoikis, a form of programmed cell death linked to cancer, has garnered significant research attention. Esophageal cancer (ESCA) ranks among the most prevalent malignant tumors and represents a major global health concern. To ascertain whether anoikis-related genes (ARGs) can accurately predict ESCA prognosis, we evaluated the predictive value and molecular mechanisms of ARGs in ESCA and constructed an optimal model for prognostic prediction. Using the Cancer Genome Atlas (TCGA)-ESCA database, we identified ARGs with differences in ESCA. ARG signatures were generated using Cox regression. A predictive nomogram model was developed to forecast ARG signatures and patient outcomes in ESCA. Gene set enrichment analysis (GSEA) was employed to uncover potential biological pathways associated with ARG signatures. Estimation of stromal and immune cells in malignant tumor tissues using expression data (ESTIMATE) and cell-type identification by estimating relative subsets of RNA transcripts analyses were used to assess differences in the immune microenvironment of the ARG signature model. Based on ARGs, the patients with ESCA were divided into high and low groups, and the sensitivity of patients to drugs in the database of genomics of drug sensitivity in cancer was analyzed. Finally, the correlation between drug sensitivity and risk score was then evaluated based on the ARG signatures. Prognostic relevance was significantly linked to the ARG profiles of 5 genes: MYB binding protein 1a (MYBBP1A), plasminogen activator, urokinase (PLAU), budding uninhibited by benzimidazoles 3, HOX transcript antisense RNA, and euchromatic histone-lysine methyltransferase 2 (EHMT2). Using the risk score as an independent prognostic factor combined with clinicopathological features, the nomogram accurately predicted the overall survival (OS) of individual patients with ESCA. Gene ontology (GO) enrichment analysis indicated that the primary molecular roles included histone methyltransferase function, binding to C2H2 zinc finger domains, and histone-lysine N-methyltransferase activity. GSEA revealed that the high-risk cohort was connected to cytokine-cytokine receptor interaction, graft-versus-host disease, and hematopoietic cell lineage, whereas the low-risk cohort was related to arachidonic acid metabolism, drug metabolism via cytochrome P450 and fatty acid metabolism. Drug sensitivity tests showed that 16 drugs were positively correlated, and 3 drugs were negatively correlated with ARG characteristic scores. Our study developed 5 ARG signatures as biomarkers for patients with ESCA, providing an important reference for the individualized treatment of this disease.

Combining clinical and molecular data for personalized treatment in acute myeloid leukemia: A machine learning approach

Background and ObjectiveThe standard of care in Acute Myeloid Leukemia patients has remained essentially unchanged for nearly 40 years. Due to the complicated mutational patterns within and between individual patients and a lack of targeted agents for most mutational events, implementing individualized treatment for AML has proven difficult. We reanalysed the BeatAML dataset employing Machine Learning algorithms. The BeatAML project entails patients extensively characterized at the molecular and clinical levels and linked to drug sensitivity outputs. Our approach capitalizes on the molecular and clinical data provided by the BeatAML dataset to predict the ex vivo drug sensitivity for the 122 drugs evaluated by the project. MethodsWe utilized ElasticNet, which produces fully interpretable models, in combination with a two-step training protocol that allowed us to narrow down computations. We automated the genes’ filtering step by employing two metrics, and we evaluated all possible data combinations to identify the best training configuration settings per drug. ResultsWe report a Pearson correlation across all drugs of 0.36 when clinical and RNA sequencing data were combined, with the best-performing models reaching a Pearson correlation of 0.67. When we trained using the datasets in isolation, we noted that RNA Sequencing data (Pearson: 0.36) attained three times the predictive power of whole exome sequencing data (Pearson: 0.11), with clinical data falling somewhere in between (Pearson 0.26). Lastly, we present a paradigm of clinical significance. We used our models’ prediction as a drug sensitivity score to rank an individual's expected response to treatment. We identified 78 patients out of 89 (88 %) that the proposed drug was more potent than the administered one based on their ex vivo drug sensitivity data. ConclusionsIn conclusion, our reanalysis of the BeatAML dataset using Machine Learning algorithms demonstrates the potential for individualized treatment prediction in Acute Myeloid Leukemia patients, addressing the longstanding challenge of treatment personalization in this disease. By leveraging molecular and clinical data, our approach yields promising correlations between predicted drug sensitivity and actual responses, highlighting a significant step forward in improving therapeutic outcomes for AML patients.

Examining molecular landscape and drug sensitivity profiles of patient-derived xenografts from resected brain metastases.

28 Background: Brain metastases (BrMet) remain an unmet need. Preclinical models of BrMet are rare but can be a valuable tool to interrogate pathology and develop novel therapeutic approaches. We previously published molecular profiles and drug sensitivity screening using novel BrMet patient-derived xenograft (PDX) models established from resected BrMet from patients with breast cancer. We further expanded our study to include BrMet from other solid tumors. Methods: We ascertained the genomic profiles of thirteen novel non-breast BrMet PDX models from patients undergoing clinically indicated surgical resection of BrMet. We then conducted drug sensitivity screening on ex vivo organoids with a panel of over 350 drugs to interrogate drug responses. We also collected information on molecular alterations and clinical treatment history from matched source patient material to correlate molecular alterations between tissue and matched PDXs when available, as well as with drug response in PDXs. Results: The PDXs included tumors that commonly (lung and melanoma) and less commonly (ovarian and sarcoma) metastasize to the brain. We observed potentially targetable DNA alterations in the DNA damaging repair pathway ( ATM, BRCA, CHEK2) and ERBB2. Using the drug sensitivity score 3 (DSS3), we assessed the sensitivity of the drugs tested. While the panel did not include the recently approved anti-HER2 targeted drugs, lapatinib (anti-HER2 TKI), and PARP inhibitors, e.g. olaparib and niraparib, were inactive against these PDXs. The three most active drugs among the BrMet PDXs were romidepsin, bortezomib, and triptolide, similar to what we observed previously for breast BrMet PDXs. When available, we correlated the DSS3 from the PDX testing with the drug exposure history and potentially actionable alterations reported in the matched source patient clinical samples. Interestingly, one case with CDKN2A/1B homozygous deletion had differential sensitivity against CDK inhibitors (highly active to dinaciclib but inactive to abemaciclib and palbociclib). Conclusions: We observed prevalent gene alterations associated with homologous recombination deficiency and HER2 among these non-breast BrMet PDX models, opening an opportunity for novel cross cancer approaches as these alterations are also reportedly linked to BrMet from breast cancer, these mechanisms may contribute to the propensity for BrMet. We also found that there are common drugs that are active across BrMet from different solid tumor types. To further explore the shared molecular features of BrMet across various solid tumors, we have established additional PDX models (near 100) from multiple solid tumors and plan to report on the molecular features.

MODL-13. ORGANOTYPIC BRAIN SLICE CULTURE PLATFORM AS A MODEL OF PEDIATRIC LOW GRADE GLIOMAS

Abstract BACKGROUND Pediatric low grade glioma (pLGG), the most common subgroup of central nervous system tumors in children, suffers from a paucity of models that support pLGG biological discovery, preclinical target identification, and drug screening studies. METHODS We propose the use of an organotypic brain slice culture (OBSC) platform as a novel model in which we are able to reliably support fresh, zero-passage pLGG patient tumor tissue for preclinical testing. We performed individualized drug screening with the frequently used up-front agents: carboplatin, vincristine, trametinib, and dabrafenib. RESULTS We have collected and established on OBSCs 100% (3/3) pLGGs since 2023. With these three patient samples, we have applied our quantitative drug screening algorithm, which combines tumor killing efficacy with healthy tissue toxicity, and yielded resultant drug sensitivity scores (DSS) to quantify patient-specific treatment efficacies. CONCLUSIONS Overall, these early results suggest that the OBSC platform may serve as a novel and reliable model for pLGG, filling a critical void in representative models for this common and often highly morbid pediatric tumor.

High-throughput neural stem cell-based drug screening identifies S6K1 inhibition as a selective vulnerability in sonic hedgehog-medulloblastoma.

Medulloblastoma (MB) is one of the most common malignant brain tumors in children. Current treatments have increased overall survival but can lead to devastating side effects and late complications in survivors, emphasizing the need for new, improved targeted therapies that specifically eliminate tumor cells while sparing the normally developing brain. Here, we used a sonic hedgehog (SHH)-MB model based on a patient-derived neuroepithelial stem cell system for an unbiased high-throughput screen with a library of 172 compounds with known targets. Compounds were evaluated in both healthy neural stem cells (NSCs) and tumor cells derived from the same patient. Based on the difference of cell viability and drug sensitivity score between normal cells and tumor cells, hit compounds were selected and further validated in vitro and in vivo. We identified PF4708671 (S6K1 inhibitor) as a potential agent that selectively targets SHH-driven MB tumor cells while sparing NSCs and differentiated neurons. Subsequent validation studies confirmed that PF4708671 inhibited the growth of SHH-MB tumor cells both in vitro and in vivo, and that knockdown of S6K1 resulted in reduced tumor formation. Overall, our results suggest that inhibition of S6K1 specifically affects tumor growth, whereas it has less effect on non-tumor cells. Our data also show that the NES cell platform can be used to identify potentially effective new therapies and targets for SHH-MB.

Immunotherapy and drug sensitivity predictive roles of a novel prognostic model in hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is one of the most significant causes of cancer-related deaths in the worldwide. Currently, predicting the survival of patients with HCC and developing treatment drugs still remain a significant challenge. In this study, we employed prognosis-related genes to develop and externally validate a predictive risk model. Furthermore, the correlation between signaling pathways, immune cell infiltration, immunotherapy response, drug sensitivity, and risk score was investigated using different algorithm platforms in HCC. Our results showed that 11 differentially expressed genes including UBE2C, PTTG1, TOP2A, SPP1, FCN3, SLC22A1, ADH4, CYP2C8, SLC10A1, F9, and FBP1 were identified as being related to prognosis, which were integrated to construct a prediction model. Our model could accurately predict patients’ overall survival using both internal and external datasets. Moreover, a strong correlation was revealed between the signaling pathway, immune cell infiltration, immunotherapy response, and risk score. Importantly, a novel potential drug candidate for HCC treatment was discovered based on the risk score and also validated through ex vivo experiments. Our finds offer a novel perspective on prognosis prediction and drug exploration for cancer patients.

Abstract P58: Utilizing Cancer Vulnerabilities and Dependencies to Explore Cancer Biomarkers by Triangulating Large-Scale Gene Knockout and Drug Response Data

Abstract The identification of biomarkers of therapeutic response for different cancers presents significant challenges owing to the heterogeneity of these diseases. Whole-genome CRISPR gene knockout screening (DepMap) and large-scale drug dosing (e.g. PRISM, and Sanger GDSC) performed across large panels of cancer cell lines reveal a whole spectrum of cancer vulnerabilities and dependencies that can be used to identify potential biomarkers and perturbed pathways in subtypes of cancers. In this study, we identified tumor type-specific dependent genes with CRISPR screening data. Functional relationships among genes were systematically examined by clustering genes with pronounced dependencies based on established biological pathways and conducting a co-dependency analysis. We found that both tumor type and subtype-specific analysis allowed the classification of cell lines based on gene dependency. Next, we correlated gene dependency of the selected dependent genes with normalised drug sensitivity scores (AUC) to triangulate potential biomarkers of drug response. We first applied this method to several tumor types which are well-characterised by known cancer drivers, and successfully identified known biomarkers and targeted pathways. However, we also identified previously unknown gene-drug relationships which will be further explored. Subsequent analysis of rarer cancers, including soft tissue sarcoma, yielded promising gene targets. This analysis pipeline can be used extensively to mine DepMap data using any subset of cell lines of interest for biomarker discovery. Citation Format: Shen Yong Yeo, Wai Yee Wong, Jesslyn, Tan Boon Toh, Valerie Shiwen Yang, Xing Yi Woo. Utilizing Cancer Vulnerabilities and Dependencies to Explore Cancer Biomarkers by Triangulating Large-Scale Gene Knockout and Drug Response Data [abstract]. In: Proceedings of Frontiers in Cancer Science; 2023 Nov 6-8; Singapore. Philadelphia (PA): AACR; Cancer Res 2024;84(8_Suppl):Abstract nr P58.

A phenotypic screening approach to target p60AmotL2-expressing invasive cancer cells

BackgroundTumor cells have the ability to invade and form small clusters that protrude into adjacent tissues, a phenomenon that is frequently observed at the periphery of a tumor as it expands into healthy tissues. The presence of these clusters is linked to poor prognosis and has proven challenging to treat using conventional therapies. We previously reported that p60AmotL2 expression is localized to invasive colon and breast cancer cells. In vitro, p60AmotL2 promotes epithelial cell invasion by negatively impacting E-cadherin/AmotL2-related mechanotransduction.MethodsUsing epithelial cells transfected with inducible p60AmotL2, we employed a phenotypic drug screening approach to find compounds that specifically target invasive cells. The phenotypic screen was performed by treating cells for 72 h with a library of compounds with known antitumor activities in a dose-dependent manner. After assessing cell viability using CellTiter-Glo, drug sensitivity scores for each compound were calculated. Candidate hit compounds with a higher drug sensitivity score for p60AmotL2-expressing cells were then validated on lung and colon cell models, both in 2D and in 3D, and on colon cancer patient-derived organoids. Nascent RNA sequencing was performed after BET inhibition to analyse BET-dependent pathways in p60AmotL2-expressing cells.ResultsWe identified 60 compounds that selectively targeted p60AmotL2-expressing cells. Intriguingly, these compounds were classified into two major categories: Epidermal Growth Factor Receptor (EGFR) inhibitors and Bromodomain and Extra-Terminal motif (BET) inhibitors. The latter consistently demonstrated antitumor activity in human cancer cell models, as well as in organoids derived from colon cancer patients. BET inhibition led to a shift towards the upregulation of pro-apoptotic pathways specifically in p60AmotL2-expressing cells.ConclusionsBET inhibitors specifically target p60AmotL2-expressing invasive cancer cells, likely by exploiting differences in chromatin accessibility, leading to cell death. Additionally, our findings support the use of this phenotypic strategy to discover novel compounds that can exploit vulnerabilities and specifically target invasive cancer cells.

Abstract 3100: Interactive analysis of drug testing data for functional precision oncology with Breeze 2.0

Abstract High throughput screening (HTS) is used to screen large numbers of small molecules in cell-based assays to predicting their potential therapeutic response and when applying approved drug libraries using HTS it becomes drug repurposing. The Drug Sensitivity and Resistance Testing (DSRT) platform established at FIMM has been extensively used in testing panels of approved and investigational drugs on cells, both conventional cell lines and primary cell models derived from cancer patient samples. The results help in deepen our understanding of drug efficacy in individual patients and across specific disease to guide future patient stratification. Breeze 2.0 is an open access, web-based platform for interactive analysis of drug screening data in real time. Breeze accepts raw data in a simple tabular format and generates detailed quality control metrics, plots, dose-response curve fits, drug response quantification in form of IC50/EC50/AUC and Drug Sensitivity Scores (DSS). Besides tabular results, Breeze generates interactive, informative, and intuitive visualizations, such as plate heatmaps, scatterplots, barplots, clustered heatmaps, circular plots, for further analysis. Based on manually-curated published datasets available through Breeze 2.0, users can instantly compare their drug response patterns with the existing DSRT profiles from patient samples, as well as cell lines. Importantly, Breeze 2.0 has integrated curated reference DSRT data tested on bone marrow samples from healthy control individuals, which plays a crucial role in understanding cancer-selective drug response by de-prioritizing generally toxic compounds. The end-users can select multiple healthy controls or relevant patient/cell line samples and choose the drugs of interest to visualize the overlaid dose responses, as well as response quantification metrics. Breeze 2.0 is built in R, PHP, MySQL and is freely accessible at https://breeze.fimm.fi/v2 Citation Format: Swapnil Potdar, Filipp Ianevski, Aleksandr Ianevski, Ziaurrehman Tanoli, Krister Wennerberg, Brinton Seashore-Ludlow, Olli Kallioniemi, Päivi Östling, Tero Aittokallio, Jani Saarela. Interactive analysis of drug testing data for functional precision oncology with Breeze 2.0 [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 3100.

Abstract 5664: Genetic alterations and pharmacological response profiles from 10 cervical cancer cell lines

Abstract Background: There are significant parallels between CC cell lines (CCCL) and newly diagnosed tumors, the most prominent being HPV infection/integrations and PI3K as well as epigenetic pathway alterations. Objectives: Identify genomic alterations associated with resistance to specific drug families in pharmacological profiles in Cell Lines (CL) and in Primary Tumors. Material and Methods: 16 public and 4 newly patient-derived CL have been extensively characterized by HPV type, genetic (WES) and protein (RPPA) profiles. Pharmacological profiling against 32 drugs was carried out in triplicate samples and repeated on 10 selected CL. Variant calling of small-scale alterations (SNVs, InDels) was performed using GATK HaplotypeCaller (v4.0.2.1). Variants annotation was performed with Annovar (v 2017/07/06) using respectively gene (refGene), known polymorphims (avsnp v147, 1000G v2015/08, ESP6500v2, ExAC) and cancer databases (COSMIC v70, ICGC v21) as well as deleterious impact predictors (dbNSFP v33). Copy number alterations were assessed using Facets (v0.5.11) with 4 blood samples from the RAIDs tumor project used to generate a “baseline” copy number profiles for CLs. Results: All 10 cell lines had strikingly similar DSS (drug sensitivity scores) to a family of four different microtubule targeting agents (MTAs). Bioinformatics subtraction of genetic alterations present in resistant (CRL 10302, CRL 7920, CRL 1550) and absent in sensitive CLs: (IC1, IC3, IC5), revealed a consistent LoF (loss of function) alteration in one single gene: PABPC3 ‘Poly A tail binding protein, cytoplasmic C3’, a gene involved in messenger RNA stability and translational initiation. Additionally, 50 gene alterations from the pathway “POST TRANSLATIONAL MODIFICATIONS OF TUBULIN” were variably detected in resistant CLs. In primary tumors from the RAIDS database, alterations in PABPC3 were present in 10/301 (3%), half of which documented by at least one database: (Cosmic, ICGC, Cancer Hotspot). All CLs were sensitive to Bortezomib, Omipalisib and Palbociclib. A large spectrum activity to Vorinostat, (HDAC inhibitor) in line with frequent (sensitizing) LoF alterations in epigenetic suppressor genes, contrasted with a selective response to other well-known CC drugs: Methotrexate, Gemcitabine, Sorafenib. Three out of CL were highly sensitive to APR246 (Eprenetapopt) as single agent. Bioinformatics gene subtraction identified OSBPL1A, a marker of synthetic lethality in the ferroptosis pathway, in all three CLs. Discussion and Conclusion: A string of LoF alterations (CSMD2/CSMD3, ZNF717, CDC27) had been previously associated with poor response to radiation or to chemo-radiation, in tumors and are present in most CLs which are derived from resistant tumors. Future objectives are to validate the present biomarker findings in the context of a clinical platform trial. Citation Format: Suzy Scholl, Pierre Gestraud, Elodie Girard, Lolita Lecompte, Maral Halladjian, Leanne de Koning, Nicolas Servant, Maud Kamal, Elaine Del Nery. Genetic alterations and pharmacological response profiles from 10 cervical cancer cell lines [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 5664.

Delineating functional and molecular impact of ex vivo sample handling in precision medicine

Consistent handling of samples is crucial for achieving reproducible molecular and functional testing results in translational research. Here, we used 229 acute myeloid leukemia (AML) patient samples to assess the impact of sample handling on high-throughput functional drug testing, mass spectrometry-based proteomics, and flow cytometry. Our data revealed novel and previously described changes in cell phenotype and drug response dependent on sample biobanking. Specifically, myeloid cells with a CD117 (c-KIT) positive phenotype decreased after biobanking, potentially distorting cell population representations and affecting drugs targeting these cells. Additionally, highly granular AML cell numbers decreased after freezing. Secondly, protein expression levels, as well as sensitivity to drugs targeting cell proliferation, metabolism, tyrosine kinases (e.g., JAK, KIT, FLT3), and BH3 mimetics were notably affected by biobanking. Moreover, drug response profiles of paired fresh and frozen samples showed that freezing samples can lead to systematic errors in drug sensitivity scores. While a high correlation between fresh and frozen for the entire drug library was observed, freezing cells had a considerable impact at an individual level, which could influence outcomes in translational studies. Our study highlights conditions where standardization is needed to improve reproducibility, and where validation of data generated from biobanked cohorts may be particularly important.

Combination drug screen identifies synergistic drug interaction of BCL-XL and class I histone deacetylase inhibitors in MYC-amplified medulloblastoma cells

PurposePatients with MYC-amplified Group 3 medulloblastoma (MB) (subtype II) show poor progression-free survival rates. Class I histone deacetylase inhibitors (HDACi) are highly effective for the treatment of MYC-amplified MB in vitro and in vivo. Drug combination regimens including class I HDACi may represent an urgently needed novel treatment approach for this high risk disease.MethodsA medium-throughput in vitro combination drug screen was performed in three MYC-amplified and one non-MYC-amplified MB cell line testing 75 clinically relevant drugs alone and in combination with entinostat. The drug sensitivity score (DSS) was calculated based on metabolic inhibition quantified by CellTiter-Glo. The six top synergistic combination hits were evaluated in a 5 × 5 combination matrix and a seven-ray design. Synergy was validated and characterized by cell counts, caspase-3-like-activity and poly-(ADP-ribose)-polymerase-(PARP)-cleavage. On-target activity of drugs was validated by immunoprecipitation and western blot. BCL-XL dependency of the observed effect was explored with siRNA mediated knockdown of BCL2L1, and selective inhibition with targeted compounds (A-1331852, A-1155463).Results20/75 drugs effectively reduced metabolic activity in combination with entinostat in all three MYC-amplified cell lines (DSS ≥ 10). The combination entinostat and navitoclax showed the strongest synergistic interaction across all MYC-amplified cell lines. siRNA mediated knockdown of BCL2L1, as well as targeted inhibition with selective inhibitors showed BCL-XL dependency of the observed effect. Increased cell death was associated with increased caspase-3-like-activity.ConclusionOur study identifies the combination of class I HDACi and BCL-XL inhibitors as a potential new approach for the treatment of MYC-amplified MB cells.Graphical abstractGraphical abstract created with BioRender.com, illustrating the workflow and summarizing main results.

Establishment of a lysosome-related prognostic signature in breast cancer to predict immune infiltration and therapy response.

Lysosomes are instrumental in intracellular degradation and recycling, with their functional alterations holding significance in tumor growth. Nevertheless, the precise role of lysosome-related genes (LRGs) in breast cancer (BC) remains elucidated. This study aimed to establish a prognostic model for BC based on LRGs. Employing The Cancer Genome Atlas (TCGA) BC cohort as a training dataset, this study identified differentially expressed lysosome-related genes (DLRGs) through intersecting LRGs with differential expression genes (DEGs) between tumor and normal samples. A prognostic model of BC was subsequently developed using Cox regression analysis and validated within two Gene Expression Omnibus (GEO) external validation sets. Further analyses explored functional pathways, the immune microenvironment, immunotherapeutic responses, and sensitivity to chemotherapeutic drugs in different risk groups. Additionally, the mRNA and protein expression levels of genes within the risk model were examined by utilizing the Gene Expression Profiling Interactive Analysis (GEPIA) and Human Protein Atlas (HPA) databases. Clinical tissue specimens obtained from patients were gathered to validate the expression of the model genes via Real-Time Polymerase Chain Reaction (RT-PCR). We developed a risk model of BC based on five specific genes (ATP6AP1, SLC7A5, EPDR1, SDC1, and PIGR). The model was validated for overall survival (OS) in two GEO validation sets (p=0.00034 for GSE20685 and p=0.0095 for GSE58812). In addition, the nomogram incorporating clinical factors showed better predictive performance. Compared to the low-risk group, the high-risk group had a higher level of certain immune cell infiltration, including regulatory T cells (Tregs) and type 2 T helper cells (Th2). The high-risk patients appeared to respond less well to general immunotherapy and chemotherapeutic drugs, according to the Tumor Immune Dysfunction and Exclusion (TIDE), Immunophenotype Score (IPS), and drug sensitivity scores. The RT-PCR results validated the expression trends of some prognostic-related genes in agreement with the previous differential expression analysis. Our innovative lysosome-associated signature can predict the prognosis for BC patients, offering insights for guiding subsequent immunotherapeutic and chemotherapeutic interventions. Furthermore, it has the potential to provide a scientific foundation for identifying prospective therapeutic targets.

Identification of efferocytosis-related subtypes in gliomas and elucidating their characteristics and clinical significance.

Introduction: Gliomas, the most prevalent tumors of the central nervous system, are known for their aggressive nature and poor prognosis. The heterogeneity among gliomas leads to varying responses to the same treatments, even among similar glioma types. In our study, we efferocytosis-related subtypes and explored their characteristics in terms of immune landscape, intercellular communication, and metabolic processes, ultimately elucidating their potential clinical implications. Methods and Results: We first identified efferocytosis-related subtypes in Bulk RNA-seq using the NMF algorithm. We then preliminarily demonstrated the correlation of these subtypes with efferocytosis by examining enrichment scores of cell death pathways, macrophage infiltration, and the expression of immune ligands. Our analysis of single-cell RNA-seq data further supported the association of these subtypes with efferocytosis. Through enrichment analysis, we found that efferocytosis-related subtypes differ from other types of gliomas in terms of immune landscape, intercellular communication, and substance metabolism. Moreover, we found that the efferocytosis-related classification is a prognostic factor with robust predictive performance by calculating the AUC values. We also found that efferocytosis-related subtypes, when compared with other gliomas in drug sensitivity, survival, and TIDE scores, show a clear link to the effectiveness of chemotherapy, radiotherapy, and immunotherapy in glioma patients. Discussion: We identified efferocytosis-related subtypes in gliomas by analyzing the expression of 137 efferocytosis-associated genes, exploring their characteristics in immune landscape, intercellular communication, metabolic processes, and genomic variations. Moreover, we discovered that the classification of efferocytosis-related subtypes has a strong prognostic predictive power and holds potential significance in guiding clinical treatment.

Novel disulfidptosis-derived gene blueprint stratifying patients with breast cancer.

Breast cancer remains the predominant cancer among females, accounting for about 24.2% of all cancer cases. Alarmingly, it is the primary cause of cancer-related mortality in women under 45. This research analyzed RNA sequencing data from 1082 TCGA-BRCA and 107 GSE58812 breast cancer patients. Single-cell RNA data from five patients in the GSE118389 data set were also studied. Using Random forest and COX regression, we developed a prognostic model. Pathway analysis employed GSVA and GO, while immune profiles were assessed via ssGSEA and MCPcounter. Mutation patterns utilized maftools, and drug sensitivity scores were derived from the GDSC database with oncoPredict. Analysis of the GSE118389 data set identified three distinct cell types: immune, epithelial, and stromal. P53 and VEGF were notably enriched. Five key genes (TMEM251, ADAMTSL2, CDC123, PSMD1, TLE1) were pinpointed for their prognostic significance. We introduced a disulfidptosis-associated score as a novel risk factor for breast cancer prognosis. Survival outcomes varied significantly between training and validation sets. Comprehensive immune profiling revealed no difference in activated CD8-positive T cells between risk groups, but a positive correlation of NK cells, neutrophils, cytotoxic lymphocytes, and monocytic cells with the riskscore was noted. Importantly, a negative association between the drug Nelarabine and riskscore was identified. This research underscores the significance of a disulfidptosis-associated gene signature in breast cancer prognosis.

The Impact of the Peptide Drug Conjugate Melflufen on the Myeloma Tumour Microenvironment

Introduction Melphalan flufenamide (melflufen), is a peptide-drug conjugate that delivers alkylating agents in tumors upon cleavage by aminopeptidases, and recently approved by the EMA for the treatment of relapsed/refractory multiple myeloma (MM). Although melflufen's MoA on tumor cells had been extensively studied, little is known about the drug impact on the tumor microenvironment. By applying single cell, multi-parametric, flow cytometry-based analysis to bone marrow mononuclear cells (BM-MNC) from MM patient samples treated ex vivo, we investigated the effect of melflufen on immune cell populations to assess for potential immune modulatory effects and to identify potential immune-based biomarkers of response. Methods Viably frozen BM-MNC from 26 samples of patients diagnosed with MM were used for multi-parametric flow cytometry-based drug sensitivity and resistance testing evaluation to melflufen after written informed consent following approved protocols in compliance with the Declaration of Helsinki. Drug sensitivity was assessed using five different concentrations of melflufen ranging between 0.5-5000 nM and cells incubated with drug or DMSO control for 72 hrs incubation. For analysis, we used antibody panels allowing us to identify and phenotype T and B lymphocytes, NK, progenitor, and malignant plasma cells (PC). Based on drug sensitivity scores (DSS) of gated PCs from drug treated and controls samples, we divided the samples into two groups - high sensitivity (HS, Q3 of PC-DSS) and low sensitivity (LS, Q1 of PC-DSS). Samples with no PCs detected (n=11) were excluded from PC subset analysis. Results The MM samples were divided based on the sensitivity of the PC population to melflufen with DSS > 21.25 defining a highly sensitive group and DSS < 15.9 defining a less sensitive group. We found that HS samples have higher abundance of mature NK and iNKT cells, although significance was low, likely due to the small cohort size. In addition, we discovered that effector memory (EM) and terminal effector (TE) T cells have higher sensitivity to melflufen, compared to naïve and central memory (CM) phenotypes (Fig. 1). More mature NK cells (Stage 5 and 6) also showed higher sensitivity than NK cells of earlier stages of maturation. Importantly, we observed that in many of the tested samples, progenitor-like CD34+CD38- cells either were not sensitive to the cytotoxic effect of the drug, or even proliferated in higher drug concentrations (Fig. 2). Interestingly, HS samples had significantly higher baseline abundance of this cell population (p<0.05). At the same time, response profiles of healthy B and malignant PCs were similar. Finally, in our tests, PCs in samples with trisomy 11 abnormality showed lower sensitivity to melflufen (p<0.05). Conclusion In a previous study, we found that effector-like cells could be associated with better response to melflufen. In this study, we observed the same trends and confirm our earlier findings demonstrating an association of NK and NKT cells with higher sensitivity to melflufen. Moreover, we identified differential response of CD4, CD8 and NK cell populations comprising the tumor microenvironment to melflufen suggesting that in addition to direct tumor cell killing, melflufen may have modulating effects on different tumor associated immune cell populations. Finally, we observed proliferation of CD34+CD38- progenitor cells, which might be relevant for lymphocyte proliferation and function, and additionally supporting potential modulating effect. Further analyses are being done to validate these findings.

A Phase 2 Study with Volasertib for Ven-HMA Relapsed/Refractory Acute Myeloid Leukemia Patients Guided By a Predictive Precision Medicine Platform

Introduction Acute myeloid leukemia (AML) represents a significant clinical challenge, with complex genetic heterogeneity that underlies the variable patient response to existing treatment modalities. While molecular features and novel therapeutic options offer promising solutions, their effectiveness is often compromised by inter-patient and intra-leukemia variability, coupled with an overall lack of biomarkers that can reliably inform therapeutic benefit. Consequently, these therapies yield clinical responses limited to only a subset of patients. Predicting a patient's response before initiating therapy optimizes treatment selection by improving the benefit/risk ratio (therapeutic index) and increasing response rates. This helps avoid treating non-responding patients, thereby reducing the occurrence of adverse effects and minimizing the costs associated with ineffective care. Here, we present an actionable strategy in AML that employs an automated, high-throughput, multi-color flow cytometry-based predictive precision medicine platform (PPMP) offering a genetic biomarker-agnostic approach to identify responders to standard-of-care (SOC) treatments and the polo-like kinase 1 (PLK-1) inhibitor volasertib. Clinical validation of the PPMP To date, we have conducted four studies evaluating the PPMP's predictive performance in myelodysplastic syndromes, adult AML, and pediatric AML with several SOC agents. For all studies, pre-induction blood or bone marrow samples were collected from patients then ex vivo treatment responses of individual leukemic cells were assessed using multi-color flow cytometry. The results were validated with the patient's actual clinical response as reported by physicians and indicated high precision-with positive predictive values (PPVs) of 83-100%. These findings help establish a PPMP-based pursuit of investigational therapies to enhance their therapeutic utility in distinct response groups. Case study: Volasertib One application of this approach involves the revival of volasertib, a PLK-1 inhibitor with demonstrated activity in AML, that was discontinued by Boehringer Ingelheim (BI) after failing its Phase 3 study. A BI Phase 2 study showed volasertib combined with low-dose cytarabine (V+LDAC) delivered a complete response (CR) + CR with incomplete count recovery (CRi) rate of 31.0% compared to 13.3% from LDAC monotherapy, and higher overall survival (OS). Similarly, in a global Phase 3 study a CR + CRi rate for V+LDAC (27.7% overall response) compared to Placebo+LDAC (17.1%) was reported. However, the Phase 3 study showed a higher frequency of fatal events with V+LDAC, especially in non-responders, resulting in lower median OS. By accurately pinpointing clinical responders, the integration of PPMP aims to identify those patients deriving clinical benefit (CR or CRi) from volasertib. Clinical trial design We plan on initiating a Phase 2a single-arm, open-label clinical trial in AML with volasertib in Ven-HMA relapsed or refractory patients, while co-developing PPMP as a companion diagnostic to selectively enroll predicted responders in a follow-up Phase 2b study (Figure 1). Intially, 20-30 patients will be enrolled. Once sufficient responders are achieved and certain performance thresholds of the drug sensitivity platform are met, an additional 20-30 patients will be enrolled, utilizing the companion diagnostic to selectively enroll predicted responders. The primary objectives will be focused on efficacy and safety of intravenous volasertib (possibly in combination with another drug) in patients with relapsed/refractory AML. Secondary objectives will include ex vivo response data collection to optimize the diagnostic parameters for the Phase 2b study. Primary Phase 2a endpoints include AML disease response assessment per pre-specified criteria. Secondary endpoints include an assessment of the diagnostic performance characteristic of Notable's drug sensitivity score in predicting in vivo volasertib clinical response. Specific strategies, including tailored dosing, robust infection control, and carefully selecting drug combination partners are designed to mitigate the adverse events observed in BI's Phase 3 trial and maximize volasertib's therapeutic impact regarding both efficacy and safety.

MODL-44. DEVELOPMENT OF A NOVEL, PERSONALIZED DRUG SENSITIVITY SCORE TO ASSESS TUMOR KILLING AND TOXICITY WITHIN AN ORGANOTYPIC BRAIN SLICE CULTURE PLATFORM

Abstract Development of new therapeutics for clinical use requires balancing tumor killing and off-target toxicity to normal tissue. We have established an organotypic brain slice culture (OBSC) platform which enables us to assess both on- and off-target drug effects via engraftment, treatment, and analysis of cell lines and uncultured patient brain tumor tissue. We have used OBSCs to test seven cell lines, ten patient tissues, and normal brain tissue against a panel of eleven different therapeutics. To determine a normalized comparison among drugs of differing potencies, we have established a drug sensitivity score (DSS) which combines both direct tumor killing via bioluminescence imaging and off-target toxicity via propidium iodide staining. Our ability to define these values independently allows us to generate normalized therapeutic window ratios for each drug-tumor-OBSC interaction. From these results, 11 different parameters, including those commonly used to define drug activity such as IC10, IC50, and area under the curve, were assessed and combined into a single ranked score which combines more than 100 data points. This algorithm enables relative comparisons among tumor responses to several drugs with less bias toward more potent compounds which produce correspondingly high off-target toxicity. Trends across DSSs for every tumor indicate that we can calculate scores throughout the entire scoring range. While we find that patient tumor tissue is generally more sensitive to treatment than cell lines, there is still a wide distribution of patient tumor sensitivities among tumor types and therapeutics. Additionally, our DSS can detect differences in tumor sensitivity after a single gene alteration in RAD18. In summary, our DSS algorithm and scoring system provides a summarized, normalized way to assess both on-target killing and off-target toxicity, allowing more comprehensive comparisons of drugs with varying potency.

MODL-03. AN EX VIVO PLATFORM BASED ON ORGANOTYPIC BRAIN SLICE CULTURE FOR INFORMING PERSONALIZED BRAIN CANCER THERAPY

Abstract While the improved classification of CNS tumors is already enhancing the scientific rigor of clinical brain cancer treatment guidelines, the impact of these integrated diagnoses based on histopathological features or molecular insight on precision oncology medicine has yet to demonstrate widespread clinical benefit. Patient tumor-derived cancer models, including patient tumor-derived cell lines, organoids, and explants, have provided practical models based on a patient’s individual tumor for effective therapeutic drugs. Importantly, patient tumor-derived organoids have shown potential to guide personalized care and optimize clinical outcomes by functionally predicting patient response to antitumor drugs prior to in vivo treatment. Our team has built an organotypic brain slice culture (OBSC)-based platform and multi-parametric algorithm that allows for the ex vivo engraftment, rapid treatment, and systemic analysis of resected patient brain tumor tissue. This OBSC platform has successfully supported engraftment of each patient tumor tested to this point, including high- and low-grade adult and pediatric tumor tissue. The established patient tumor-derived organoids show the ability to activate the endogenous astrocytes within OBSCs, which have been reported in brain tumor. Whole exome sequencing analysis showed that patient tumor tissue engrafted onto OBSCs maintained a significant genetic resemblance to their parent tumor, while tumor tissue expanded in vitro displayed a distinctly different DNA profile. Additionally, our algorithm calculates dose-response relationships of both tumor kill and normal brain tissue toxicity, generating summarized drug sensitivity scores based on therapeutic window and allowing us to normalize response profiles across a panel of FDA-approved and exploratory agents. Summarized patient tumor scores after OBSC treatment revealed positive associations to clinical outcomes, suggesting the OBSC platform can provide rapid, accurate, functional testing to ultimately guide patient care.

Computational Approach Based on Independent Drug Action Identifies Clinically Relevant Combinations for Acute Myeloid Leukemia

Background Acute myeloid leukemia (AML) is clinically and genetically heterogenous due to complex co-mutation landscape. Combination therapy has shown effectiveness in overcoming tumor heterogeneity and acquired resistance across multiple cancers including AML. However, the discovery of novel combinations has traditionally relied on cell line-based measurements that often fail clinically due to inter- and intra-tumor heterogeneity of patients. To address this challenge of finding clinically relevant combination therapy for AML, we applied the independent drug action (IDA) principle (Palmer & Sorger, 2017) to computationally predict drug combinations using ex vivo monotherapy responses measured in primary AML patient specimens (n > 1000). Methods We utilized the ex vivo drug sensitivity testing (DST) of primary AML samples from three studies; 1. FPMTB (Malani et al. 2021), 164 patient responses to 515 single agents; 2. Beat-AML (Tyner et al. 2018), 579 patient responses to 166 single agents; 3. Oregon Health & Science University (OSHU), 596 patient responses to 51 single agents and 170 2-drug combinations (Figure 1A). To summarize the multiparametric dose-response relationship into a single metric, drug sensitivity score (DSS) (Yadav et al. 2014) was calculated from the dose-response cell viability data of the samples using the BREEZE pipeline (Potdar et al. 2020) to derive monotherapy efficacy. We modified IDACombo package (Ling & Huang, 2020) to predict the mean combination DSS of every pairwise combination. IDAComboscore was calculated for each predicted combination to determine if it is more efficacious than its constituent monotherapies. The top hits from predicted combinations were selected by filtering for combinations with mean DSS in the top 10 th percentile, and positive IDAComboscore. These combinations were then ranked in descending order of IDAComboscore. For validation, Spearman correlation was applied to compare the predicted DSS against ex vivo DST-derived DSS of 170 combinations from the OSHU dataset. Results By applying IDA-based prediction on FPMTB and Beat-AML datasets, we identified that the mean combination DSS exceeded monotherapy DSS (p < 2.2x10 -6, Wilcoxon rank-sum test), suggesting that the predicted combination would outperform monotherapy. After accounting for monotherapies that displayed pan-toxicity across healthy samples and primary AML samples (mean DSS > 95 th percentile), we found venetoclax as the top combination partner of the top 100 most effective predicted combinations (FPMTB, n = 33; Beat-AML, n = 35). We further interrogated the makeup of venetoclax-based combinations and found kinase inhibitors (PI3Ki, MEKi, mTORi), and epigenetic modifiers (BETi, HDACi) as common partners. A total of 8 (FPMTB) and 18 (Beat-AML) out of these 100 combinations were made of FDA-approved drugs (both partners). Next, we investigated the robustness of the predictions across 2 datasets, and we found 1767 overlapping combinations with a strong correlation between their predicted mean DSS (Spearman ρ = 0.74, p < 2.2x10 -6). We finally validated the performance of IDA predictions using ex vivo DST data of 170 unique combinations from 596 primary AML samples. We found that the predicted DSS values of the 170 combinations strongly correlated with the actual ex vivo DST-derived combination DSS (Spearman ρ = 0.76, p < 2.2x10 -6, Figure 1B). Using 75 th percentile of DSS distribution as effective combination cutoff, the approach showed an accuracy of 84.7%. We then asked if disease status of patients affected prediction accuracy and found that IDA approach successfully predicted effective combinations for patients at diagnosis as well as at relapse (diagnosis ρ = 0.74 vs relapse ρ = 0.73). Conclusion We have computationally identified and validated clinically actionable drug combinations for AML using ex vivo monotherapy responses from >1000 tumor specimens based on IDA principle. Irrespective of pharmacological interaction, IDA-based combinations are more efficacious than single drugs in AML primary tumors. Our approach enables the major advantage of identifying effective combinations for the entire patient population as an alternative to personalized combinations. Further associations with genetic makeup and transcriptomics signature can be leveraged to identify suitable combinations.