Author List:(presenter in bold, rest alpha): Sam DeLuca, Tom Howd, Fontina Kelley, Mike DaSilva, Tim Desmet, Katie Larkin, Tamara Mason, Steve Ferreira, Jim Meldrim, Maegan Harden, Alyssa Macbeth, Betty Woolf, Kathleen Tibbetts, Eric Banks, Laura Gauthier, Sam Friedman, Lee Lichtenstein, Wendy Brodeur, Matt Lebo, Sheila Dodge, Heidi Rehm, Stacey Gabriel, Niall Lennon.

Broad Institute of MIT and Harvard.

As precision medicine moves from the realm of clinical utility studies towards standard of care, there is a pressing need to decrease turnaround times for WGS data generation as part of an end-to-end clinical workflow. In the last year, we have validated and launched a clinical WGS product with a deliverable of 95% of bases at ≥ 20X coverage within the industry standard of 28 calendar days. Here, we present how we have since built a scalable workflow around this clinical grade, gold standard PCR-Free genome process that allows for sample intake to return of data in 5 days.  

In order for us to accomplish our goal of a 5 day clinical grade genome, we needed to take advantage of a streamlined, dynamic workflow which incorporates our internally developed LIMS and state of the art automation technology. Within 48 hours of sample intake, we are capable of registering blood tubes into our LIMS system, performing an automated extraction followed by a series of internal QC assays, generating PCR-Free libraries, and having them on sequencing instruments. We have implemented several principles from lean manufacturing and dynamic work design in order to minimize waste and eliminate unnecessary touch points or delays in the workflow.

For this rapid WGS application, we have chosen to utilize an S2 flow cell on the NovaSeq 6000, which provides complete sequencing data in only 25 hours. Data transfer from the sequencing platform to an analysis pipeline is automatically started upon run completion. Significant improvements to our data analysis pipeline have allowed for us to integrate GATK calling with advanced single sample filtering to provide vcf files that include functional annotations within 48 hours of sequencing being completed.

Michelle Cipicchio, Gina Vicente, Sophie Low, Matthew DeFelice, Mark Fleharty, Jonna Grimsby, Hayley Lyon, Alyssa Macbeth, Justin Abreu, Brendan Blumenstiel, Maura Costello, Tim Desmet, Niall Lennon, Stacey Gabriel

The Broad Institute Genomics Platform has developed a complete end-to-end redesign of our targeted sequencing products, enabling lower input, faster turnaround time, lower cost, and improved sequencing quality. The new unified workflow can accommodate a variety of input DNA sample types, including cell-free, FFPE, as well as low and high quality genomic DNA.Our dual-indexed library construction process offers an option for duplex unique molecular identifiers (UMIs) for applications requiring detection of low allele fraction variants in deep coverage sequencing projects. A more cost-effective library construction option with lower coverage and longer inserts is also available for germline and population genetics focused research projects, to maximize coverage while minimizing sequencing costs.The hybrid selection workflow is compatible with multiple targeted panels of varying sizes, including our redesigned and improved exome panel, which starts with the Twist Human Core Exome design and incorporates ~2 Mb of additional custom content curated from both our germline and somatic users. This customized exome covers ACMG59 genes, and now also targets additional RefSeq and OMIM (Online Mendelian Inheritance in Man) putative gene sequences, COSMIC (Catalogue of Somatic Mutations in Cancer) variants, key promoters and other motifs that have been identified as potential cancer hot spots, and the complete mitochondrial genome. This new exome product shows improved evenness compared to our previous version, enabling libraries to achieve deeper coverage with less sequencing.  The selection workflow, which utilizes IDT’s xGen Hybridization and Wash kit, improves quality while reducing turnaround time. This workflow also provides the flexibility to customers to choose panels from any vendor, such as IDT and Twist Bioscience. In addition to enabling large scale research projects, our exome and a subset of targeted panels will be available for clinical sequencing in our CLIA-licensed, CAP-accredited facility.  

Jonna Grimsby1, Matthew De Felice1, Mark Fleharty1, Madeleine Duran1, Michelle Cipicchio1, Katie Larkin1, Nicholas Fitzgerald1, Michael Nasuti1, Justin Abreu1, Maura Costello1, Tom Howd1, Carrie Cibulskis1, Niall Lennon1, Brendan Blumenstiel1

1Broad Institute of MIT and Harvard 320 Charles St Cambridge, MA 02141

Cell-free DNA (cfDNA) (from whole blood liquid biopsy) sequencing and tumor mutation detection will enable the detection and monitoring of cancer status to direct effective patient-specific treatments. When blood plasma is collected for cfDNA extraction, lysis of white blood cells can contaminate the cfDNA with genomic DNA (gDNA), limiting the number of cfDNA molecules available for library construction. In combination with previous sensitivity benchmarking studies and through a series of liquid biopsy projects, we sought to understand the relationship between cfDNA input into NGS library construction, gDNA contamination, achievable duplex depth (read depth based on UMIs [unique molecular identifiers]) and sensitivity for calling low allele fraction variants. Cell-free DNA was extracted from 64 patients and NGS libraries were constructed using duplex UMI adapters. Target regions were selected through hybridization and capture using array-synthesized probes, libraries were sequenced to ~20,000-50,000x raw read depth, reads were deduplicated by UMIs, and duplex consensus reads were generated. When duplex UMI consensus coverage was plotted as a function of cfDNA input (ng) into library construction, a few samples with the highest cfDNA input were observed to have low duplex depth. We observed that an “inflated” cfDNA extraction yield in combination with a low library yield can be indicative of gDNA contamination in our sample, resulting in lower duplex depth. In previous benchmarking studies, we have shown that as raw sequencing depth increases, duplex coverage increases, but duplex coverage reaches a plateau once no new molecules (UMI families) are being observed, resulting in a measurable limit of sensitivity for detecting mutations. Understanding this relationship between actual cfDNA input, raw sequencing depth, duplex depth, and sensitivity is critical for assessing sequencing requirements and for evaluating the true limitations and capabilities of liquid biopsy sequencing assays.

Jacqueline Dion, Alyson Reardon1, Michelle Cipicchio1,  Katie Larkin1, Stacey Gabriel1, Niall Lennon1

1Broad Institute of MIT and Harvard 320 Charles St Cambridge, MA 02141

Through direct to patient cancer studies, over 2000 saliva samples have been collected at the Broad for genomic analysis thus far.  Samples are received daily through direct to patient engagement. Relative ease of a saliva submission paired with the direct to patient model has removed many barriers to patient involvement in research. Count Me In (www.mbcproject.org, www.ascproject.org, https://mpcproject.org) research initiatives are now using this pipeline to collect saliva and a solid tumor biopsy to sequence participants routinely. Saliva collection tubes can be shipped to participants and easily dropped in a mailbox to be returned.

DNA collection through blood requires meeting with a trained phlebotomist and flash freezing the blood. Fresh blood can also be used however, it has a shorter shelf life than Oragene™ saliva collection tubes. Saliva will stay intact for years at room temperature and lysis can occur safely in the collection tube with minimal intervention. DNA from saliva is generally high molecular weight, and the percent of bacterial contamination is on average far less than from other methods of saliva collection.

Optimization of saliva extraction methods is important in order to obtain substantial yield and purity from the sample. Seven experimental designs were investigated using similar input and a designated elution volume. Four saliva sample pools were used in triplicate for each experiment performed. Evaluation of the workflow, timing, throughput capacity, purity, and yield helped to identify the best extraction method. From this research, the extraction has been brought to scale in an automated high throughput fashion that requires little manual intervention. The Broad Institute utilized the expertise in bringing a manual process to scale to realize a workflow that can digitally track chain of custody of each sample and meet the high capacity of the whole genome sequencing pipeline.

Datasheet for the Clinical Whole Genome Sequencing product