Rence probes were located in genomic regions with low frequency copy number changes. The hybridization and ligations were carried out as per instructions and fragment analysis was performed on an ABIPRISM 3130xl capillary sequencer. The data were visualized using peak scanner v1.0 software and the exported data was analyzed with Coffalyser software v8 (MRCHolland, Amsterdam, the Netherlands). Calculation of signal ratios was carried out as described by Mistry et al. [38]. Stringent criteria were adopted for data analysis using Coffalyser software and experiments were repeated twice for reproducibility.Study populationTumor tissues were collected from pancreatic cancer patients during surgery between January 2002 and September 2009, snapfrozen in liquid nitrogen directly after resection and subsequently stored at 280 uC. A total of 171 tumor tissues, that contained at least 10 tumor by H E staining were analyzed in the present study. The clinical and histopathological characteristics of the patients are given in Table 1. The cell lines A549, SW1116, SW620, HS766T, MiaPaCa and LoVo were commercially obtained from DprE1-IN-2 manufacturer American Type Culture collection (ATCC) [36,37].Statistical analyses Histopathological assessment of cellular composition of tissue biopsiesThree different tissue sections were selected randomly for hematoxylin and eosin (H E) staining and histological validation. get Fruquintinib Slides were scanned with the ScanScope GL System (Aperio Technologies, Vista, CA, USA) and visualized using the ImageScope Software. For each tissue sample, three pathologists evaluated independently the histology and percentage of normal, tumor and stroma cells (Figure S1). Only samples with more than 10 tumor cells were pursued further. Of 171 1317923 tumors that were analyzed for mutations, 163 were malignant and 8 non-malignant tumors. Of the 
163 patients with malignant tumors, survival data were available for 153 patients, of whom 150 patients had malignant tumors of pancreatic origin including ductal adenocarcinomas (n = 135) and rare carcinomas (n = 15). The rest (n = 3) were carcinoma of ampulla of Vater (Table 1). The Kaplan eier method was employed to determine the cumulative survival curves using time period (in months) between date of operation and the date of death. Differences between the groups were analyzed by the log-rank test. Univariate and multivariate Cox regression analyses were used to determine proportional hazard ratios. For multivariate analysis variables included were gender, age at surgery, TNM status, tumor differentiation grade and histological status of tumors. All statistical analyses were carried out by using SASH version 9.2 (SAS Institute Inc., Cary, NC).Genomic DNA extractionFrozen pancreatic tissue samples were individually cut into 20 mm thick slices with a cryotome Leica CM 1850 UV at 234 uC. The tissue slices were covered with liquid nitrogen and gently ground by three turns with a micropestle made of polypropylene (Eppendorf, Hamburg, Germany) that fitted into 2 ml Eppendorf tubes. DNA from tissue slices and from cell lines was extracted using the AllPrep Isolation Kit (Qiagen, Hilden, Germany). DNA 
from cell lines with known KRAS mutations in codon 12, 13 and 61 were used as controls that included, A549 cell line with G12S (GGT.AGT) mutation; MiaPaCa, G12C (GGT.TGT); SW 1116, G12A (GGT.GCT); SW 620 G12V (GGT.GTT); LS-174, G12D (GGT.GAT); LoVo, G13D (GGC.GAC); HS 766T, Q61H (CAA.CAC). DNA samples from healthy controls were included as.Rence probes were located in genomic regions with low frequency copy number changes. The hybridization and ligations were carried out as per instructions and fragment analysis was performed on an ABIPRISM 3130xl capillary sequencer. The data were visualized using peak scanner v1.0 software and the exported data was analyzed with Coffalyser software v8 (MRCHolland, Amsterdam, the Netherlands). Calculation of signal ratios was carried out as described by Mistry et al. [38]. Stringent criteria were adopted for data analysis using Coffalyser software and experiments were repeated twice for reproducibility.Study populationTumor tissues were collected from pancreatic cancer patients during surgery between January 2002 and September 2009, snapfrozen in liquid nitrogen directly after resection and subsequently stored at 280 uC. A total of 171 tumor tissues, that contained at least 10 tumor by H E staining were analyzed in the present study. The clinical and histopathological characteristics of the patients are given in Table 1. The cell lines A549, SW1116, SW620, HS766T, MiaPaCa and LoVo were commercially obtained from American Type Culture collection (ATCC) [36,37].Statistical analyses Histopathological assessment of cellular composition of tissue biopsiesThree different tissue sections were selected randomly for hematoxylin and eosin (H E) staining and histological validation. Slides were scanned with the ScanScope GL System (Aperio Technologies, Vista, CA, USA) and visualized using the ImageScope Software. For each tissue sample, three pathologists evaluated independently the histology and percentage of normal, tumor and stroma cells (Figure S1). Only samples with more than 10 tumor cells were pursued further. Of 171 1317923 tumors that were analyzed for mutations, 163 were malignant and 8 non-malignant tumors. Of the 163 patients with malignant tumors, survival data were available for 153 patients, of whom 150 patients had malignant tumors of pancreatic origin including ductal adenocarcinomas (n = 135) and rare carcinomas (n = 15). The rest (n = 3) were carcinoma of ampulla of Vater (Table 1). The Kaplan eier method was employed to determine the cumulative survival curves using time period (in months) between date of operation and the date of death. Differences between the groups were analyzed by the log-rank test. Univariate and multivariate Cox regression analyses were used to determine proportional hazard ratios. For multivariate analysis variables included were gender, age at surgery, TNM status, tumor differentiation grade and histological status of tumors. All statistical analyses were carried out by using SASH version 9.2 (SAS Institute Inc., Cary, NC).Genomic DNA extractionFrozen pancreatic tissue samples were individually cut into 20 mm thick slices with a cryotome Leica CM 1850 UV at 234 uC. The tissue slices were covered with liquid nitrogen and gently ground by three turns with a micropestle made of polypropylene (Eppendorf, Hamburg, Germany) that fitted into 2 ml Eppendorf tubes. DNA from tissue slices and from cell lines was extracted using the AllPrep Isolation Kit (Qiagen, Hilden, Germany). DNA from cell lines with known KRAS mutations in codon 12, 13 and 61 were used as controls that included, A549 cell line with G12S (GGT.AGT) mutation; MiaPaCa, G12C (GGT.TGT); SW 1116, G12A (GGT.GCT); SW 620 G12V (GGT.GTT); LS-174, G12D (GGT.GAT); LoVo, G13D (GGC.GAC); HS 766T, Q61H (CAA.CAC). DNA samples from healthy controls were included as.