Objectives: In the current work, we propose to use gene expression analysis of tumor biopsies obtained sequentially before and after treatment, to discover new surrogate biomarkers of therapeutic benefit.Such strategy was applied in a Phase Ⅰ study that combines Sorafenib (S), a multikinase inhibitor, with Dacarbazine (D), a cytotoxic agent.Methods: This study was performed in accordance with protocols approved by the ethics committee.We sequentially obtained 2 biopsies from17 patients includes in this trial and harbouring accessible advanced solid tumors.One biopsy was collected at baseline before treatment, and the other one was collected at day 21 corresponding to the end of the first cycle.Among those 17 patients, 10 display a therapeutic benefit (PFS>3 month) whereas 7 had a progressive disease (PFS<3months).Gene expression was analyzed, using dual-color labelling, with 44k pangenomic 60-mer arrays (Agilent TechnologiesTM).Direct comparison was performed for each biopsies couple, in duplicate, with dye-swap hybridization.We selected genes differentially expressed in each paired biopsies then we compared expression profiles between patients with or without therapeutic benefit.By the use of paired samples, we minimized inter-individual variability and we preserved a relevant statistical power despite the low number of patients.The analysis was based on leave-one-out strategy where we generated 17 different learning sets by excluding one sample and using the 16 remaining to define genes associated to therapeutic benefit (gene filtering at P<0.001 and Student test).The biological significance analysis was performed with Ingenuity(R) Pathway knowledge database.Results: We obtained 17 different gene sets of 21 to 81 genes.Their union generated a 247 genes set that clustered efficiently the 17 patients according to the therapeutic benefit status.The intersection of the 17 gene sets correspond to only 4 genes (NM_172229, NM_003129, NM_006403 and AI613259).Interestingly, those 4 genes expression allowed to distinguish patients according to their therapeutic benefit status.Surprisingly, the gene expression of the 7 known Sorafenib target were not modified.However, the biological pathway analysis among the 247 genes set reveals that 15 genes are related to Sorafenib biological activity: RAF1 pathway (8 genes); KDR pathway (4 genes), KIT pathway (2genes) and PDGFRβ pathway (1 gene).Conclusion: In the present study, we show the potential value of genomic analysis in phase Ⅰ trial.Despite the high heterogeneity of tumors histology treated and the limited number of patients, we demonstrate that gene expression profiling of biopsies sequentially obtained is a powerful approach to discover putative biomarkers in early clinical trials.The biomarkers discovered in this study will deserve further validation.Nevertheless, they would be useful to predict therapeutic benefit as soon as the end of the first cycle of the S and D combination.This study also confirms the feasibility of the genomic strategy based on sequential biopsies in terms of samples size, RNA extraction and early detection of gene differentially expressed in response to chemotherapy.