Skip to main content
Premium Trial:

Request an Annual Quote

UPenn's Center for Personalized Diagnostics Launches MiSeq Cancer Panels, Studies Utility


The Center for Personalized Diagnostics at the University of Pennsylvania Perelman School of Medicine has launched two next-gen sequencing-based gene panels for cancer patients and is studying their utility compared to single-gene tests as well as exome and genome sequencing.

The center, founded last year, is a collaboration between the Department of Pathology and Laboratory Medicine, where it is currently housed, and UPenn's Abramson Cancer Center. Early this summer, the CPD will move into a larger space that will enable it to expand.

Its mission is to provide physicians in the University of Pennsylvania Health Care System with genomic diagnostic cancer tests to help them make treatment decisions, with the goal being to extend these services in the future to the Penn Cancer Network – a group of community hospitals in Pennsylvania, New Jersey, and Delaware that collaborate with the Abramson Cancer Center.

The CPD is currently equipped with two Illumina MiSeqs, a HiSeq 2500, an Ion Torrent PGM, and Affymetrix Cytoscan array technology. Earlier this year, the center's clinical laboratory was CLIA-certified and CAP-accredited, and in mid-February it launched two cancer sequencing panels, a hematologic malignancy panel that covers 33 genes and a solid tumor sequencing panel that includes 47 genes.

Right now, the CPD offers testing for leukemias and lymphomas, and brain, melanoma, and lung tumors, but the plan is to expand it to other tumor types over the next several months. "We really wanted to release this to a smaller group initially, so we can refine our processes and make sure everything was working well," Robert Daber, the technical director of clinical genomics at the center, told Clinical Sequencing News.

The center's current setup allows it to sequence and analyze about 1,000 samples per year, but as it adds more tumor types, it expects to expand to be able to handle several thousand samples per year.

The center validated both panels on about 700 samples last year. The solid tumor panel is based on Illumina's off-the-shelf TruSeq Amplicon Cancer Panel, except for one gene that did not perform well during validation. The hematologic panel is a custom TruSeq Amplicon panel, and both panels run on the MiSeq. Some of the genes included in the panels are sequenced in full while for others only hotspot areas are covered.

Turnaround time for the test is approximately 14 days, which the lab hopes to reduce to 7 to 10 days when its testing volume increases and it performs more than one MiSeq run per week, which can fit between 30 and 40 patient samples. The quickest turnaround time to date – for an urgent leukemia case where an answer was required as quickly as possible – has been three days.

Prior to designing and testing the panels, the CPD performed a head-to-head comparison between several target enrichment methods, including Agilent's HaloPlex and Ion Torrent's AmpliSeq, as well as between the MiSeq and PGM.

At the time, they decided that the MiSeq and TruSeq amplification "worked best for us," Daber said, but the performance of all the technologies is "a moving target," and for an updated version of its panels that the lab is currently developing, "we are going to put them all head to head again, because we feel they have all made significant improvements," he said.

For the data analysis, the CPD has developed its own software, allowing it to find insertions and deletions up to about 30 base pairs in length, with a sensitivity down to 5 percent allele frequency, which Daber said is better than any commercial informatics solutions.

Also, during the validation phase, the lab developed quality control parameters for the input DNA. Based on the quality of the DNA, the researchers adjust the detection sensitivity of their bioinformatics software to account for the higher background error in low-quality samples, a feature that Daber said is somewhat unique to their lab.

To annotate somatic variants, the center built an "extensive knowledgebase" from information in public databases for the genes it analyzes. This required "cleaning up" databases of germline variants, for example dbSNP, which are contaminated with somatic mutations, and databases of somatic mutations, such as COSMIC, which also contain germline polymorphisms, Daber said.

The clinical reports that go to physicians contain the mutations with references from the literature, but no direct links to drugs that could target them. "It's the physicians who are educated about knowing what the drugs are," Daber explained. "We don't make any drug recommendations."

During its first seven weeks of service, the lab took on 76 patient samples and returned 64 reports. Most of these were FFPE samples, while about a dozen were blood or bone marrow samples. In 52, they found at least one disease-associated mutation, four patients had variants of unknown significance, three FFPE tumor samples were too degraded to perform the assay, and five samples yielded no mutations.

Because the tests are so new, it is not known yet how many clinicians have used the results to make treatment decisions, but the CPD has an ongoing outcomes study to answer this question that will be completed in a year and a half.

Over the first six months, the NGS panels will be run in parallel with existing single-gene diagnostic tests that are covered by the panel, "so we can capture information about the added benefit of next-gen panels over the standard single-gene approaches," Daber said.

"By doing this study, we are able to actually define where each of the technologies is best. Because next-gen [sequencing] is just another technology, and it doesn't answer all the questions."

Besides providing clinical testing, the CPD is also developing new tests and diagnostic technologies and comparing different types of tests. A couple of pilot projects, for example, address the utility of targeted panel sequencing, exome sequencing, and genome sequencing in different tumor cohorts in terms of their diagnostic yield. Initially, these projects will involve about 50 exomes, and a smaller number of genomes, but those numbers might increase if it turns out early on that exome or genome sequencing is more useful than panels.

Another project aims to detect copy number alterations in sequence data from gene panels in FFPE samples, which could potentially replace SNP arrays for that application. "FFPE SNP arrays are really hard and difficult assays," Daber said. "But for sequencing FFPE samples, we now have a protocol that works very robustly."

The CPD also has a number of ongoing collaborations to develop new genomic technologies, such as single-cell genomics and circulating tumor DNA analysis. The goal is to develop these into clinical assays eventually, and the CPD is providing its next-gen sequencing and informatics expertise to the projects. "By participating in these research studies, we can really design them in such a way that it's easier to translate them into a clinical test," Daber said.

Among the greatest challenges so far in providing NGS-based diagnostic tests has been the existing infrastructure. "It's a very disruptive technology [that] changes the status quo in a lot of different areas," Daber said.

For example, most electronic medical records are not yet able to capture multi-analyte genomic data. Also, billing for NGS tests is difficult because "there was no paradigm for this type of testing established yet," he said, and the center is in discussions with payors about billing. Finding the right personnel with expertise in next-gen sequencing in a clinical setting, or bioinformatics and biology, is also "a bottleneck for the field," he added.

In the meantime, doctors at UPenn are starting to adopt the panels. "They knew about this testing long before we offered it," Daber said, and provided feedback on what genes to include and what type of information would be most valuable to them.