« Back to Farmacia Journal 4/2026

COMPARATIVE ANALYSIS OF DRUG RESPONSE METRICS ACROSS TUMOUR TYPES UNDER CAMPTOTHECIN EXPOSURE

LINA CHUB ¹*, WASHINGTON BARBOSA DA SILVA ², FELIPE DE OLIVEIRA LOPES CAVALCANTI ³, RONNI GERALDO GOMES DE AMORIM ¹

Download Full Article PDF

The LN(IC₅₀) and AUC metrics are standard in cancer pharmacogenomics, but their interchangeability remains poorly understood. We tested the hypothesis that a high correlation between these metrics does not necessarily imply identical results in the binary classification of tumour cell line sensitivity. Using GDSC2 data from 116 cell lines (BLCA, BRCA, CESC, GBM), a correlation analysis was performed, and the frequency of discordance between LN(IC₅₀) and AUC for camptothecin was evaluated. Robustness was assessed by varying classification thresholds and comparing RMSE between concordant and discordant observations. Despite a high correlation (ρ = 0.967), discordance reached 8.6%. Alternative thresholds reduced discordance to 6.9%. RMSE differences were not significant (p = 0.0808), although the discordant group was small. Discordant observations clustered near the classification boundaries were observed (p < 0.001), which is to be expected when two highly correlated measures are divided at the median. The results demonstrate that highly correlated continuous metrics do not guarantee that binary outcomes will be identical, and emphasise the value of jointly analysing several metrics of drug response when interpreting borderline cases.