5959. Pipeline of Bioinformatics Analysis The bioinformatics analysis began with the sequencing data that was generated from HiSeq2000. In the first step, the adapter sequence in the raw data was removed, and low quality reads that had too many unreadable or low quality bases were discarded, producing “clean data”. For the second step, a Burrows-Wheeler Aligner was used 
for alignment. BWA can generate results in a BAM file format, which is a requirement for a number of the subsequent processes, such as fixing the mate information of the alignment, adding read group information, and removing duplicate reads caused by PCR. After these processes, the final BAM files used for the variant calling were prepared. Single nucleotide polymorphism analysis was performed using SAMtools, and 1268798 web potential somatic single nucleotide variants were predicted using Varscan . We then used our filter pipeline to identify somatic mutations, with the following major criteria: The adjacent somatic mutation distance should be equal to or greater than 10 bp; the mapping quality score should not be significantly lower than 30; the base quality score should not be significantly lower than 20; there should be a significant allele frequency change between the tumor and the matched adjacent normal tissue; the mutation should not be in gap-aligned reads; mutations should not be significantly enriched within 5 bp of the 59 or 39 ends of the read; and mutations should not be in a simple repeat region. Small insertion/ deletions were detected using SAMtools, and structure variants and copy number variants were identified using BreakDancer and a method we devised by ourselves based on the Segseq algorithm, respectively. ANNOVAR was used to annotate confident variant results. The final variant scan could then be used in the downstream advanced analysis pipeline. QC was applied throughout the entire process to obtain clean data, alignment and called variants. Library Construction Libraries of qualified genomic DNA were prepared for pairedend analysis on the Illumina Cluster Station and Illumina HiSeq 2000. To minimize the likelihood of systematic bias in sampling, three paired-end libraries with an insert size of 500 bp were prepared for all samples in this study. The qualified genomic DNA was fragmented 16963441 into lengths of approximately 500 bp using Covaris. The double-stranded DNA fragments contained 39 or 59 overhangs, which were converted into blunt ends using T4 DNA polymerase and Klenow enzyme. Adenosine was added to the 39 end of the blunt phosphorylated DNA fragments, and adapters were then ligated to both ends. The correctly ligated products were separated by agarose gel electrophoresis and then purified using the QIAquick gel extraction kit. The purpose of this step is to remove residual free and self-ligated adaptors and to select correctly sized templates for cluster generation. DNA fragments with adapter molecules on both ends were 20719936 selected and amplified. Polymerase chain reaction was performed with the two primers that annealed to the ends of the adapters. The number of PCR cycles was minimized to avoid skewing the representation of the library, and the PCR products were checked and purified by agarose gel electrophoresis. They were then used in the library test step and the average fragment size and molar concentration of the library were determined by using an Agilent 2100 Bioanalyzer and an ABI Real-Time PCR System, respectively. Genomic Profiling We aligned sequ
