CellMinerCDB is a resource that simplifies access and exploration of cancer cell line pharmacogenomic data across different sources

Database and query tool designed for cancer research community to facilitate integration and study of molecular and pharmacological data for the NCI-60 cancerous cell lines. The NCI-60, panel of 60 diverse human cancer cell lines used by the Developmental Therapeutics Program of the U.S. National Cancer Institute to screen over 100,000 chemical compounds and natural products since 1990.

These are cancer cell line pharmacogenomics datasets that work with rcellminer and CellMinerCDB (see publications for details).

An example for extracting data from the rcellminer and CellMinerCDB compatible packages:

library(nciSarcomaData)

# DRUG DATA ----
drugAct <- exprs(getAct(nciSarcomaData::drugData))
drugAnnot <- getFeatureAnnot(nciSarcomaData::drugData)[["drug"]]

# MOLECULAR DATA ----
## List available datasets
names(getAllFeatureData(nciSarcomaData::molData))

## Extract data and annotations
expData <- exprs(nciSarcomaData::molData[["exp"]])
mirData <- exprs(nciSarcomaData::molData[["mir"]])

expAnnot <- getFeatureAnnot(nciSarcomaData::molData)[["exp"]]
mirAnnot <- getFeatureAnnot(nciSarcomaData::molData)[["mir"]]

# SAMPLE DATA ----
sampleAnnot <- getSampleData(nciSarcomaData::molData)

SclcCellMinerCDB is a resource that simplifies access and exploration of cancer cell line pharmacogenomic data across different sources

Array-based comparative genomic hybridization (aCGH) is a powerful technique for detecting gene copy number variation. It is generally considered to be robust and convenient since it measures DNA rather than RNA. In the current study, we combine copy number estimates from four different platforms (Agilent 44 K, NimbleGen 385 K, Affymetrix 500 K and Illumina Human1Mv1_C) to compute a reliable, high-resolution, easy to understand output for the measure of copy number changes in the 60 cancer cells of the NCI-DTP (the NCI-60). We then relate the results to gene expression. We explain how to access that database using our CellMiner web-tool and provide an example of the ease of comparison with transcript expression, whole exome sequencing, microRNA expression and response to 20,000 drugs and other chemical compounds. We then demonstrate how the data can be analyzed integratively with transcript expression data for the whole genome (26,065 genes). Comparison of copy number and expression levels shows an overall medium high correlation (median r = 0.247), with significantly higher correlations (median r = 0.408) for the known tumor suppressor genes. That observation is consistent with the hypothesis that gene loss is an important mechanism for tumor suppressor inactivation. An integrated analysis of concurrent DNA copy number and gene expression change is presented. Limiting attention to focal DNA gains or losses, we identify and reveal novel candidate tumor suppressors with matching alterations in transcript level.

Transcript levels for 46,834 isoforms (fragments per kilobase per million reads, FPKM)

This is the repository of data and scripts for the analysis of the CellminerCDB-SCLC manuscript and website (https://discover.nci.nih.gov/SclcCellMinerCDB/)

CellMiner-SCLC (https://discover.nci.nih.gov/SclcCellMinerCDB) integrates 118 patient-derived cell lines with drug sensitivity and genomic datasets, including high resolution methylome and RNAseq data. CellMiner-SCLC provides a new resource for SCLC research for this “recalcitrant cancer”. Of fundamental importance, we demonstrate the reproducibility and stability of the cell line datasets from different institutions (CCLE, GDSC, CTRP, NCI and UTSW). We validate the classification based on four master transcription factors: NEUROD1, ASCL1, POU2F3 and YAP1 and show transcription networks connecting them with the MYC genes (MYC, MYCL1 and MYCN) and the NOTCH and HIPPO pathways. We find that the 4 subsets express specific surface markers for antibody-targeted therapies. The YAP1-driven (SCLC-Y) cell lines differ from the other subsets by expressing the NOTCH pathway, epithelial-mesenchymal-transition (EMT) and antigen-presenting machinery (APM) genes, and by responding to mTOR and AKT inhibitors, suggesting the potential of NOTCH modulators, YAP1 inhibitors and immune checkpoint inhibitors for SCLC-Y tumors.

Gene composite transcript levels for 23,826 genes (fragments per kilobase per million reads, FPKM)

Additional file 1: Supplementary Figure 1: DDR subtyping analysis overview. A. DDR network overview. HR: Homologous recombination. MMEJ: microhomology-mediated end-joining. NHEJ: Non-homologous end-joining. BER: Base excision repair. NER: Nucleotide excision repair. MMR: Mismatch repair. DR: Direct reversal repair. TLS: Translesion synthesis. Checkpoint: Damage sensing and signaling. FA: Fanconi Anemia. Numbers in parentheses indicate number of pathway genes analyzed by our method. B. WE score formula and Essentiality Scaling Factor (ESF) criteria. Supplementary Figure 2: DDR cluster prevalence and DDR pathway single gene heatmaps. A. DDR cluster prevalence in GEMINI cohort. B. DDR cluster prevalence in IMPOWER133 cohort. C. GEMINI DDR pathway single gene expression heatmap. D. IMPOWER133 DDR pathway single gene expression heatmap. Supplementary Figure 3: IMpower133 DDR cluster differential gene expression and quantitative set analysis (QuSAGE) BG signature results. Supplementary Figure 4: SCLC CellMiner cell line DDR optimal number of k clusters elbow plot. Supplementary Figure 5: SCLC CellMiner cell line DDR cluster prevalence and marker expression. A. DDR cluster prevalence in SCLC cell line models. B. DNA damage responsive transcription factors, intra-S cell cycle checkpoint, and G2/M cell cycle checkpoint machinery expression across SCLC cell line DDR clusters. C. SLFN11 protein expression in SCLC cell line DDR clusters. D. Total RB1 and RB1-S807.S811 phospho protein expression across SCLC cell line DDR clusters. E. Cell line DDR cluster MYC family gene expression. Supplementary Figure 6: SCLC cell line DDR cluster mutation profiling. Oncoprints for TP53, RB1, and DDR gene mutations in DDR Low, Intermediate, and High clusters. Mutation data is from CCLE whole exome sequencing [17]. Supplementary Figure 7: SCLC cell line DDR cluster nonsynonymous tumor mutational burden (TMB). TMB data is from CCLE whole exome sequencing [17]. Supplementary Figure 8: SCLC/hgNEC PDX/CDX DDR gene expression and cell cycle state distribution. A. Expression of DNA damage responsive transcription factors, intra-S, and G2/M cell cycle checkpoint effectors in PDX/CDX DDR clusters. B. PDX/CDX DDR cluster scRNAseq cell cycle state distributions. Supplementary Figure 9: IMpower133 subtyping method comparison. A. Three way alluvial plot demonstrating assignment overlap between MDACC SCLC subtypes [5], Genentech (GNE) subtypes [7], and DDR status. B. IMpower133 DDR cluster and GNE subtype assignment table. Values listed represent the number of samples in each assignment category. The corresponding X2 Pearson residual dot plot comparing DDR cluster and GNE subtype assignments is shown on the right. Supplementary Figure 10: GEMINI and IMpower133 DDR cluster neuroendocrine score single gene heatmaps. A. GEMINI DDR cluster neuroendocrine score single gene expression heatmap. B. IMpower133 DDR cluster neuroendocrine score single gene expression heatmap. Supplementary Figure 11: SCLC CellMiner DDR cluster neuroendocrine features. A. SCLC cell line DDR cluster NE scores. B. SCLC cell line DDR cluster neuroendocrine status as reported by Tlemsani et al. C. SCLC PDX/CDX DDR cluster NE scores. Supplementary Figure 12: IMpower133 SCLC-A only immune checkpoint marker expression. Boxplots depict mRNA expression of CD274/PD1, CTLA4, and HAVCR2/TIM3 in SCLC-A tumors, split by their DDR status. Supplementary Figure 13: SCLC/hgNEC PDX/CDX DDR cluster MHC Class I scRNAseq expression. Supplementary Figure 14: IMpower133 DDR cluster chemotherapy response analysis. A. RECIST Best Overall Response (BOR) for IMpower133 DDR clusters following frontline EP chemotherapy. PD: Progressive disease. SD: Stable disease. PR: Partial response. CR: Complete response. B. Progression free survival Kaplan Meier plot for SCLC-A DDR clusters following frontline EP chemotherapy. C. Forest plot for progression free survival for SCLC-A DDR clusters following frontline EP chemotherapy. Supplementary Figure 15: IMpower133 DDR cluster all subtypes chemotherapy response analysis. A. Overall survival (OS) Kaplan Meier plot for all subtypes DDR clusters following frontline EP chemotherapy. B. Progression free survival (PFS) Kaplan Meier plot for all subtypes DDR clusters following frontline EP chemotherapy. C. Forrest plots for all subtype DDR cluster OS and PFS outcomes following frontline EP chemotherapy. Supplementary Figure 16: MDACC GEMINI subtype switching plots. Subtype space and SCLC-DMC subtype calls for MDACC GEMINI subtype switching patients. Supplementary Figure 17: MDACC GEMINI subtype switching patient outcomes following frontline chemoimmunotherapy. A. Overall survival outcomes following frontline chemoimmunotherapy. B. Progression free survival following frontline chemoimmunotherapy. Supplementary Figure 18: Promoter methylation changes for MDACC GEMINI subtype switching patients from baseline to progression following frontline chemoimmunotherapy. For all panels, data presented is RRBS methylation data previously published by Heeke et al [14]. Supplementary Figure 19: Circulating tumor DNA fraction for subtype switching tumors at baseline and progression following frontline chemoimmunotherapy. Supplementary Figure 20: SC53 Leiden cluster specific top drivers of plasticity. scRNAseq data for SC53 is from Gay et al. [5]. Supplementary Figure 21: WebGestalt over-representation analysis of SC53 Leiden Cluster 8 marker genes. Supplementary Figure 22: SC53 Leiden cluster specific random walk simulations. scRNAseq data for SC53 is from Gay et al. [5]. Supplementary Table 1: WE score individual DDR gene Essentiality Scaling Factor (ESF) assignments and example raw WE score generation. Supplementary Table 2: SCLC cell line whole exome sequencing mutation enrichment statistics by DDR cluster. Genes not listed in Supplementary Table 2 were not mutated in any DDR Low, DDR Intermediate, or DDR High cell line models. Supplementary Table 3: SC53 dynamic genes by Leiden cluster.

Background: Afatinib is an irreversible epidermal growth factor receptor tyrosine kinase inhibitor, and it plays a role in hepatocellular carcinoma (LIHC). This study aimed to screen a key gene associated with afatinib and identify its potential candidate drugs.Methods: We screened afatinib-associated differential expressed genes based on transcriptomic data of LIHC patients from The Cancer Genome Atlas, Gene Expression Omnibus, and the Hepatocellular Carcinoma Database (HCCDB). By using the Genomics of Drug Sensitivity in Cancer 2 database, we determined candidate genes using analysis of the correlation between differential genes and half-maximal inhibitory concentration. Survival analysis of candidate genes was performed in the TCGA dataset and validated in HCCDB18 and GSE14520 datasets. Immune characteristic analysis identified a key gene, and we found potential candidate drugs using CellMiner. We also evaluated the correlation between the expression of ADH1B and its methylation level. Furthermore, Western blot analysis was performed to validate the expression of ADH1B in normal hepatocytes LO2 and LIHC cell line HepG2.Results: We screened eight potential candidate genes (ASPM, CDK4, PTMA, TAT, ADH1B, ANXA10, OGDHL, and PON1) associated with afatinib. Patients with higher ASPM, CDK4, PTMA, and TAT exhibited poor prognosis, while those with lower ADH1B, ANXA10, OGDHL, and PON1 had unfavorable prognosis. Next, ADH1B was identified as a key gene negatively correlated with the immune score. The expression of ADH1B was distinctly downregulated in tumor tissues of pan-cancer. The expression of ADH1B was negatively correlated with ADH1B methylation. Small-molecule drugs panobinostat, oxaliplatin, ixabepilone, and seliciclib were significantly associated with ADH1B. The protein level of ADH1B was significantly downregulated in HepG2 cells compared with LO2 cells.Conclusion: Our study provides ADH1B as a key afatinib-related gene, which is associated with the immune microenvironment and can be used to predict the prognosis of LIHC. It is also a potential target of candidate drugs, sharing a promising approach to the development of novel drugs for the treatment of LIHC.

BackgroundGliomas are the most common intracranial nervous system tumours that are highly malignant and aggressive, and mitochondria are an important marker of metabolic reprogramming of tumour cells, the prognosis of which cannot be accurately predicted by current histopathology. Therefore, Identify a mitochondrial gene with immune-related features that could be used to predict the prognosis of glioma patients.MethodsGliomas data were downloaded from the TCGA database and mitochondrial-associated genes were obtained from the MITOCARTA 3.0 dataset. The CGGA, kamoun and gravendeel databases were used as external datasets. LASSO(Least absolute shrinkage and selection operator) regression was applied to identify prognostic features, and area and nomograms under the ROC(Receiver Operating Characteristic) curve were used to assess the robustness of the model. Single sample genomic enrichment analysis (ssGSEA) was employed to explore the relationship between model genes and immune infiltration, and drug sensitivity was used to identify targeting drugs. Cellular studies were then performed to demonstrate drug killing against tumours.ResultsCOX assembly mitochondrial protein homolog (CMC1), Cytochrome c oxidase protein 20 homolog (COX20) and Cytochrome b-c1 complex subunit 7 (UQCRB) were identified as prognostic key genes in glioma, with UQCRB, CMC1 progressively increasing and COX20 progressively decreasing with decreasing risk scores. ROC curve analysis of the TCGA training set model yielded AUC (Area Under The Curve) values >0.8 for 1-, 2- and 3-year survival, and the model was associated with both CD8+ T cells and immune checkpoints. Finally, using cellMiner database and molecular docking, it was confirmed that UQCRB binds covalently to Amonafide via lysine at position 78 and threonine at position 82, while cellular assays showed that Amonafide inhibits glioma migration and invasion.ConclusionOur three mitochondrial genomic composition-related features accurately predict Survival in glioma patients, and we also provide glioma chemotherapeutic agents that may be mitochondria-related targets.

Table S7. The aasociation analysis between USP36 levels and gefitinib activity in NSCLC cells according to CellMiner database.

Fas apoptosis inhibitory molecule 2 (FAIM2) is an important member of the transmembrane BAX inhibitor motif containing (TMBIM) family. However, the role of FAIM2 in tumor prognosis and immune infiltration has rarely been studied. Here, we conducted a pan-cancer analysis to explore the role of FAIM2 in various tumors and further verified the results in glioma through molecular biology experiment. FAIM2 expression and clinical stages in tumor samples and para-cancerous samples were analyzed by TIMER2 database, GEPIA database, and the TISIDB database. The role of FAIM2 on prognosis was analyzed via GEPIA2. We utilized the ESTIMATE algorithm to evaluate the ImmuneScore and StromalScore of various tumors. In addition, we explored the correlation between FAIM2 expression and tumor immune cell infiltration by the TIMER2 database. The immune checkpoint genes, tumor mutation burden (TMB), microsatellite instability (MSI), mismatch repair (MMR), and DNA methylation related to FAIM2 were analyzed based on the TCGA database. The correlation between FAIM2 expression with Copy number variations (CNV) and methylation is explored by GSCA database. Protein-Protein Interaction (PPI) analysis was obtained from the STRING database and the CellMiner database was used to explore the association between FAIM2 expression and drug response. FAIM2 co-expression genes were studied by the LinkedOmics database. Immunohistochemistry, Western Blotting Analysis, Cell Viability Assay, Colony Formation Assay, and Edu staining assay were used in the molecular biology experiments section. The FAIM2 expression was down-regulated in most tumors and highly expressed FAIM2 was associated with a better prognosis in several cancers. FAIM2 plays an essential role in the tumor microenvironment and is closely associated with immune Infiltration in various tumors. The expression of FAIM2 was closely correlated to TMB, MSI, MMR, CNV, and DNA methylation. Furthermore, FAIM2 related genes in the PPI network and its co-expression genes in glioma are involved in a large number of immune-related pathways. Molecular biology experiments verified a cancer suppressor role for FAIM2 in glioma. FAIM2 may serve as a potential pan-cancer biomarker for prognosis and immune infiltration, especially in glioma. Moreover, this study might provide a potential target for tumor immunotherapy.

Additional file 11. Table S2: Drug sensitivity analysis of HDACs based on the CellMiner database.

BackgroundThe widespread use of immune checkpoint inhibitors (anti-CTLA4 or PD-1) has opened a new chapter in tumor immunotherapy by providing long-term remission for patients. Unfortunately, however, these agents are not universally available and only a minority of patients respond to them. Therefore, there is an urgent need to develop novel therapeutic strategies targeting other co-inhibitory molecules. However, comprehensive information on the expression and prognostic value of co-inhibitory molecules, including co-inhibitory receptors and their ligands, in different cancers is not yet available.MethodsWe investigated the expression, correlation, and prognostic value of co-inhibitory molecules in different cancer types based on TCGA, UCSC Xena, TIMER, CellMiner datasets. We also examined the associations between the expression of these molecules and the extent of immune cell infiltration. Besides, we conducted a more in-depth study of VISTA. ResultThe results of differential expression analysis, correlation analysis, and drug sensitivity analysis suggest that CTLA4, PD-1, TIGIT, LAG3, TIM3, NRP1, VISTA, CD80, CD86, PD-L1, PD-L2, PVR, PVRL2, FGL1, LGALS9, HMGB1, SEMA4A, and VEGFA are associated with tumor prognosis and immune cell infiltration. Therefore, we believe that they are hopefully to serve as prognostic biomarkers for certain cancers. In addition, our analysis indicates that VISTA plays a complex role and its expression is related to TMB, MSI, cancer cell stemness, DNA/RNA methylation, and drug sensitivity.ConclusionsThese co-inhibitory molecules have the potential to serve as prognostic biomarkers and therapeutic targets for a broad spectrum of cancers, given their strong associations with key clinical metrics. Furthermore, the analysis results indicate that VISTA may represent a promising target for cancer therapy.

Each tab of the excel file is one transcriptome. For 9 organisms from Brawand et al. [35], 20% of genes with lowest GC were considered as Gini genes. (XLSX)

Background: The TLC Domain Containing 1 (TLCD1) protein, a key regulator of phosphatidylethanolamine (PE) composition, is distributed across several cellular membranes, including mitochondrial plasma membranes. Existing research has revealed the impact of TLCD1 on the development of non-alcoholic fatty liver disease. However, there remains a gap in comprehensive pan-cancer analyses of TLCD1, and the precise role of TLCD1 in cancer patient prognosis and immunological responses remains elusive. This study aims to provide a comprehensive visualization of the prognostic landscape associated with TLCD1 across a spectrum of cancers, while shedding light on the potential links between TLCD1 expression within the tumor microenvironment and immune infiltration characteristics.Methods: TLCD1 expression data were obtained from GTEx, TCGA, and HPA data repositories. Multiple databases including TIMER, HPA, TISIDB, cBioPortal, GEPIA2, STRING, KEGG, GO, and CancerSEA were used to investigate the expression pattern, diagnostic and prognostic significance, mutation status, functional analysis, and functional status of TLCD1. In addition, we evaluated the relationship between TLCD1 expression and immune infiltration, tumor mutational burden (TMB), microsatellite instability (MSI), and immune-related genes in pan-cancer. Furthermore, the association of TLCD1 with drug sensitivity was analyzed using the CellMiner database.Results: We found that TLCD1 is generally highly expressed in pan-cancers and is significantly associated with the staging and prognosis of various cancers. Furthermore, our results also showed that TLCD1 was significantly associated with immune cell infiltration and immune regulatory factor expression. Using CellMiner database analysis, we then found a strong correlation between TLCD1 expression and sensitivity to anticancer drugs, indicating its potential as a therapeutic target. The most exciting finding is that high TLCD1 expression is associated with worse survival and prognosis in GBM and SKCM patients receiving anti-PD1 therapy. These findings highlight the potential of TLCD1 as a predictive biomarker for response to immunotherapy.Conclusion: TLCD1 plays a role in the regulation of immune infiltration and affects the prognosis of patients with various cancers. It serves as both a prognostic and immunologic biomarker in human cancer.

Background: The polypyrimidine tract-binding protein (PTBP) nuclear ribonucleoprotein family of proteins, including PTBP1, PTBP2 and PTBP3, regulate the process of cell proliferation, differentiation, apoptosis and carcinogenesis. PTBPs exhibit oncogenic effects in certain tumors. However, the role of PTBPs in pan-cancer remains unclear. Our study examined the clinical significance and mechanism of PTBPs in pan-cancer.Methods: We compared the expression of PTBPs in paired and unpaired tissue samples from the Cancer Genome Atlas (TCGA) database. Univariate and multivariate Cox regression, Kaplan–Meier curves, and time-dependent receiver operating characteristic (ROC) curves were used to assess the prognostic significance of PTBPs in pan-cancer. The cBioPortal database also identified genomic abnormalities in PTBPs. TISIDB, TCGA, and Cellminer were used to investigate the relationship between PTBP expression and immune subtypes, immune checkpoint (ICP) genes, tumor mutational burden (TMB), microsatellite instability (MSI), tumor-infiltrating immune cells, and chemosensitivity. cBioPortal was used to search for PTBP co-expressing genes in pan-cancer, and GO and KEGG enrichment analyses were performed to search for PTBP-related signaling pathways.Results:PTBPs were shown to be widely upregulated in human tumor tissues. PTBP1 showed good prognostic value in ACC, KIRP, and LGG; PTBP2 in ACC and KICH; and PTBP3 in ACC, LGG, and PAAD, with AUC >0.7. PTBPs were differentially expressed in tumor immune subtypes and had a strong correlation with tumor-infiltrating lymphocytes (TILs) in the tumor microenvironment (TME). In addition, PTBP expressions were related to ICP, TMB, and MSI, suggesting that these three PTBPs may be potential tumor immunotherapeutic targets and predict the efficacy of immunotherapy. Enrichment analysis of co-expressed genes of PTBPs showed that they may be involved in alternative splicing, cell cycle, cellular senescence, and protein modification.Conclusion: PTBPs are involved in the malignant progression of tumors. PTBP1, PTBP2 and PTBP3 may be potential biomarkers for prognosis and immunotherapy in pan-cancer and may be novel immunotherapeutic targets.

D Correlation between gene expression and basal cell adhesion to VN and reduction in cell adhesion upon Plg treatment. Adhesion 4 h after seeding the cells on VN (expressed as cell index) was measured in the experiment described in panel C. The extent of cell adhesion reduction was calculated at the indicated time point in panel C as percentage of the cell index value measured in control wells treated with vehicle. Both parameters were correlated with the expression levels of the indicated genes. Gene expression data were downloaded from the CellMiner web tool. The table shows gene name, Pearson correlation coefficient (r) and p-value.. List of tagged entities: ITGAV (ncbigene:3685), ITGB3 (ncbigene:3690), PLAU (ncbigene:5328), PLAUR (ncbigene:5329), SERPINE1 (ncbigene:5054), PLG (uniprot:P00747), microarray,xCELLigence

Over the last decade the next-generation sequencing and ‘omics techniques have become indispensable tools for medicine and drug discovery. These techniques have led to an explosion of publicly available data that often goes under-utilized due to the lack of bioinformatic expertise and tools to analyze that volume of data. Here, we demonstrate the power of applying two novel computational platforms, the NCI’s CellMiner Cross Database and Lantern Pharma’s proprietary artificial intelligence (AI) and machine learning (ML) RADR® platform, to identify biological insights and potentially new target indications for the acylfulvene derivative drugs LP-100 (Irofulven) and LP-184. Analysis of multi-omics data of both drugs within CellMinerCDB generated discoveries into their mechanism of action, gene sets uniquely enriched to each drug, and how these drugs differed from existing DNA alkylating agents. Data from CellMinerCDB suggested that LP-184 and LP-100 were predicted to be effective in cancers with chromatin remodeling deficiencies, like the ultra-rare and fatal childhood cancer Atypical Teratoid Rhabdoid Tumors (ATRT). Lantern’s AI and ML RADR® platform was then utilized to build a model to test, in silico, if LP-184 would be efficacious in ATRT patients. In silico, RADR® aided in predicting that, indeed, ATRT would be sensitive to LP-184, which was then validated in vitro and in vivo. Applying computational tools and AI, like CellMinerCDB and RADR®, are novel and efficient translational approaches to drug discovery for rare cancers like ATRT.