AbstractsMedical & Health Science

Introduction: There is unmet clinical need for automated data processing of deep sequencing in heterogeneous sample populations in molecular oncology specimens. We have developed clinical assays using Next Generation Sequencing (NGS) for deep sequencing in both solid tumor and hematological indications. We have targeted five genes (ASXL1, RUNX1, EZH2, ETV6, and TP53) involved in myelodysplastic syndrome (MDS) and three genes (BRAF, c-KIT, NRAS) in late stage melanoma. We designed and implemented a data processing pipeline that integrates novel and commercial analysis software to provide automated rules-based filtering and annotation of variants. Methods: We used a targeted resequencing approach with the Fluidigm Access Array system for library creation and molecular barcoding for sample multiplexing. Samples were processed in duplicate to filter out potential false positives. Libraries were sequenced using the Ion Torrent Personal Genome Machine (PGM) (Life Technologies, Carlsbad, CA) and the Illumina MiSeq (Illumina Inc., San Diego, CA) systems. Raw data in FastQ format was preprocessed using Windows batch commands and Perl programs that were integrated with NextGENe (SoftGenetics, State College, PA) analysis software for quality trimming, alignment and variant calling against NCBI curated references. Variant calls were compared between duplicate samples and annotated using a separate Perl program that queried databases containing known germline and somatic variants. Results: Quality control was implemented by trimming all downstream reads that contained specified Q-scores (Phred, Q16 or above) as well as requiring a minimum coverage depth of 500X per exon. Only variant calls confirmed at greater than 5% in both duplicate samples and with a balanced ratio of forward and reverse reads were considered reportable in a clinical setting. Variants appearing in only one of the duplicates were consistent with systematic errors. Variant calls were annotated with existing database information or were saved in a database of novel mutations by the annotation pipeline. The pipeline removes all systematic errors in known cell line samples by filtering metrics. Conclusions: We have developed a robust and automated pipeline for NGS data analysis in the clinical lab that is independent of sequencing platform. The implementation of an automated processing pipeline increases the efficiency and consistency of analysis. Hands-on time for data analysis has been decreased from 30 minutes per sample to less than 5 minutes per sample. Carrying duplicates through the entire multiplexing and sequencing process also helps reduce test based false-positive errors (e.g. PCR errors) while robust analytical metrics remove other sources of error (e.g. strand bias). We have found an average of 6-10 false positive calls in both sample types, all of which were filtered out by the pipeline presented here