Background: In the era of personalized cancer treatment, patient selection to enrich for specific targets is key to improve treatment outcome. Given the genomic instability of tumors in time we addressed whether analysis of the primary tumor or the metastasis is most suitable for patient selection for targeted treatment. Methods: With next generation (SOLiD 3+ system, Life Technologies) and targeted sequencing (DNA capture-arrays, Agilent) we have analyzed the genetic environment of a cancer mini-genome–the exons of 1,265 cancer-associated genes- of the primary tumor and its sequential liver metastasis in 21 colorectal cancer (CRC) patients (pts). Using laser microdissection >80% tumor cell-enriched DNA was isolated from in total 47 samples, including healthy tissue of 5 pts. Sequence data was aligned and mapped against the human reference genome (GRCh37/hg19) using BWA. To limit the influence of SNV (Single Nucleotide Variation) caller artifacts we applied 3 different published methods (our PERL script, VarScan and GATK toolkit) and considered the overlapping variants as reliable. To predict the possible impact of identified SNV's on protein function polymorphism phenotyping version2 was used. Results: All 47 samples were successfully sequenced (median coverage=87x, range: 52-217). In total 6,696 known (84%) and 1,305 novel (16%) SNV’s were identified. These included 1,460 non-synonymous and 29 gained/lost stop-codons, potentially altering protein function. On average 83 potentially relevant SNV's were gained in the metastasis and 70 SNVs present in the primary tumor were lost. Interestingly, these genetic variations affected several essential pathways. This demonstrated that tumor evolution caused losses and gains of critical genes in time. In this sample set genetic selection pressure concerning the amount of gained/ lost SNV's on the metastasis due to chemotherapy could not be shown. Conclusions: These results indicate that metastatic disease significantly differs from its primary origin in CRC pts. This suggests that for the selection of the appropriate patient population for targeted therapy the metastasis better reflects the actual disease status.