CellMiner: a web-based suite of genomic and pharmacologic tools to explore transcript and drug patterns in the NCI-60 cell line set

Authors

Reinhold WC

Sunshine M

Liu H

Varma S

Kohn KW

Morris J

Doroshow J

Pommier Y

Doi

10.1158/0008-5472.CAN-12-1370

PMID: 22802077 · DOI: 10.1158/0008-5472.CAN-12-1370 · Journal: Cancer Research (2012)

TL;DR

This paper introduces CellMiner, a web-based bioinformatics suite for exploring and integrating transcript expression (22,217 genes, 360 microRNAs) and drug activity (18,549 compounds including 91 FDA-approved drugs) across the NCI-60 cell line panel. The tools provide z-score-normalized expression patterns, quality-controlled drug sensitivity data, and a pattern comparison engine that identifies correlations between gene expression, microRNA levels, and compound activity without requiring bioinformatics expertise.

Cohort & data

  • Cell line panel: NCI-60 (60 human cancer cell lines spanning 9 tissue types) from cellline_nci60.
  • Transcript data: Integration of 5 microarray platforms (Affymetrix HG-U95, HG-U133, HG-U133 Plus 2.0, GeneChip Human Exon 1.0 ST; Agilent Whole Human Genome Oligo Microarray) covering 22,217 genes.
  • MicroRNA data: 360 microRNAs from Agilent Human miRNA Microarray V2.
  • Drug activity data: GI50 values for 18,549 compounds (from 47,540 tested) from the NCI Developmental Therapeutics Program (DTP), determined by sulphorhodamine B assay at 48 hours.

Key findings

  • Z-score integration across 5 transcript platforms enables composite gene expression patterns with built-in quality control (probe range filtering at 1.2 log2, Pearson correlation thresholds of r >= 0.30 and r >= 0.60 for probe retention).
  • ABCB1 overexpression in NCI-ADR-RES and HCT-15 cell lines correlates with doxorubicin resistance, consistent with P-glycoprotein-mediated drug efflux.
  • The Pattern Comparison tool with CDH1 as input retrieves known transcriptional repressors ZEB1 (r = -0.63), SNAI2 (r = -0.47), ZEB2 (r = -0.37), and TWIST1 (r = -0.47), validating the tool’s biological relevance.
  • Using erlotinib (NSC 718781) as pattern comparison input, gefitinib and lapatinib ranked 5th and 6th among 18,549 compounds, and afatinib ranked 3rd, demonstrating the tool’s ability to identify drugs with shared mechanisms of action targeting EGFR.
  • A colon-specific input pattern identified TRIM15 (r = 0.901), RNF43 (ubiquitin ligase upregulated in colon cancer), and VIL1 (known colon cancer diagnostic marker) as top correlated genes.
  • The miR-17-92 oncogenic cluster members showed high inter-correlation (r = 0.77-0.96) and MYC correlation (r = 0.61), consistent with MYC transcriptional regulation of this cluster.

Genes & alterations

  • ABCB1 — Overexpression associated with multidrug resistance; P-glycoprotein efflux of doxorubicin and romidepsin.
  • EGFR — Target of erlotinib, gefitinib, lapatinib, and afatinib; drug activity pattern comparison example.
  • CDH1 — E-cadherin; negatively regulated by ZEB1, SNAI2, ZEB2, TWIST1; positively correlated with miR-200 family.
  • MYC — Transcriptional activator of the miR-17-92 cluster; correlated (r = 0.61) with miR-18a across NCI-60.
  • TOP1 — Previously validated using CellMiner tools for novel gene regulation mechanisms.
  • CHEK2 — Genomic and proteomic analyses using CellMiner revealed genetic inactivation or endogenous activation across NCI-60.
  • ZEB1 — Transcriptional repressor of CDH1 (r = -0.63); EMT regulator.
  • RNF43 — Ubiquitin ligase upregulated in colon cancer; colon-specific expression in NCI-60 (pattern comparison).

Clinical implications

  • CellMiner enables identification of drugs with similar mechanisms of action (e.g., EGFR inhibitor class clustering), potentially guiding combination therapy or alternative agent selection.
  • Colon-specific drug pattern matching identified selumetinib (AZD6244) as a potential colorectal cancer therapy and sunitinib as a candidate (in clinical trial for colorectal cancer at time of publication).
  • ABCB1 expression patterns can predict multidrug resistance to substrates including doxorubicin and romidepsin.

Limitations & open questions

  • The NCI-60 is a cell line panel, not patient tumors; in vivo microenvironment, heterogeneity, and immune interactions are not captured.
  • Drug activity based on GI50 at 48 hours may not reflect clinical pharmacokinetics or long-term efficacy.
  • Pattern correlations are observational (Pearson’s); causality cannot be inferred without functional validation.
  • The tool requires NSC numbers for drug queries, which may limit accessibility for clinical researchers unfamiliar with the DTP numbering system.
  • No correction for multiple comparisons in the “significant correlations” output (p < 0.05 threshold without FDR correction).

Citations from this paper used in the wiki

  • “NCI-ADR-RES and HCT-15 are the two cell lines most resistant to doxorubicin in the NCI-60, which is consistent with the fact they over-express ABCB1, whose gene product P-gp pumps the drug from the cell.”
  • “The two other FDA-approved EGFR-targeted drugs gefitinib and lapatinib ranked 5th and 6th, and afatinib (BIBW2992), which is in advanced clinical trials and also targets EGFR-ERB kinase ranked 3rd among the 18,549 drugs.”
  • “Using the pattern comparison tool with CDH1 as input, one finds the transcriptional repressors ZEB1 (TCF8), SNAI2 (SLUG), ZEB2 (SIP1), and TWIST1 each has significant negative correlation to CDH1 transcript levels, at −0.63, −0.47, −0.37, and −0.47, respectively.”
  • “MYC ranked 37th in the gene list with a correlation coefficient of 0.61. The MYC correlation is consistent with the fact that the miR-17-92 cluster is a transcriptional MYC target.”

This page was processed by paper-compiler on 2026-05-06.