Generation of a Predictive Melphalan Resistance Index by Drug Screen of B-Cell Cancer Cell Lines

BACKGROUND Recent reports indicate that in vitro drug screens combined with gene expression profiles (GEP) of cancer cell lines may generate informative signatures predicting the clinical outcome of chemotherapy. In multiple myeloma (MM) a range of new drugs have been introduced and now challenge conventional therapy including high dose melphalan. Consequently, the generation of predictive signatures for response to melphalan may have a clinical impact. The hypothesis is that melphalan screens and GEPs of B-cell cancer cell lines combined with multivariate statistics may provide predictive clinical information. MATERIALS AND METHODS Microarray based GEPs and a melphalan growth inhibition screen of 59 cancer cell lines were downloaded from the National Cancer Institute database. Equivalent data were generated for 18 B-cell cancer cell lines. Linear discriminant analyses (LDA), sparse partial least squares (SPLS) and pairwise comparisons of cell line data were used to build resistance signatures from both cell line panels. A melphalan resistance index was defined and estimated for each MM patient in a publicly available clinical data set and evaluated retrospectively by Cox proportional hazards and Kaplan-Meier survival analysis. PRINCIPAL FINDINGS Both cell line panels performed well with respect to internal validation of the SPLS approach but only the B-cell panel was able to predict a significantly higher risk of relapse and death with increasing resistance index in the clinical data sets. The most sensitive and resistant cell lines, MOLP-2 and RPMI-8226 LR5, respectively, had high leverage, which suggests their differentially expressed genes to possess important predictive value. CONCLUSION The present study presents a melphalan resistance index generated by analysis of a B-cell panel of cancer cell lines. However, the resistance index needs to be functionally validated and correlated to known MM biomarkers in independent data sets in order to better understand the mechanism underlying the preparedness to melphalan resistance.