At Michigan Medicine, WGS becomes the backbone of molecular pathology

When Annette S. Kim, MD, PhD, arrived at the University of Michigan in 2023, the molecular pathology laboratory was doing what many high-performing academic medical centers do: making a complex system work through expertise, discipline, and a lot of human effort.

The lab’s germline workflow was built around an exome (targeted) panel. It was familiar. It produced results. But it was also labor-intensive, limited in scale, and increasingly failed to answer the research questions.

Michigan Medicine was sending out samples for exomes and genomes, but it wasn’t ideal. 

“For our division with an active research mission, outsourcing could potentially disperse data and expertise that Michigan wanted to retain and build on internally,” says Kim, director of the division of Diagnostic Genetics and Genomics at the University of Michigan. “Bringing sequencing in-house would consolidate that data under institutional stewardship, keep interpretive and analytical know-how in the lab, and create a foundation the team could continually re-analyze as research priorities shifted.”

Kim and her team considered several paths: move first to an exome platform, send out sequencing while keeping interpretation internal, or build a whole-genome sequencing workflow in-house. They chose the most ambitious route.

Choosing that route meant the lab could not simply adopt an off-the-shelf solution; the team had to research and develop much of the surrounding infrastructure itself. That included engineering liquid-handling automation for scale, building informatics and analysis workflows, and re-architecting how variant data was stored and interpreted.

“There was really no reason for us to be sending anything out,” Kim says. “There was no reason to do sort of a stepwise validation. We could just expeditiously move straight to germline and then report out exome slice panels, exomes, or genomes from just a single test on a whole genome backbone.”

The change was not simply a technology upgrade. It was a redesign of the lab’s operating model.

Before the transition, the workflow for the targeted panel required two full days of wet-bench work, totaling roughly 16 to 20 full-time employee hours per run. The work was physically demanding, Kim says, leaving little time for breaks. Also, it was a slow turnaround. The kit could run only eight  samples at a time, creating bottlenecks as volume increased and requiring multiple technologists to be assigned to the bench.

Sequencing added another 30 hours, followed by a full day of mapping and pipeline steps that, for years, had to be started manually. Some automation had been introduced near the end of the platform’s life, but the pipeline still required hands-on intervention.

Coverage was another challenge. Some genes included in Michigan’s exome slice panels were not covered by the kit at all. Others were technically present but had poor coverage, sometimes less than 10x even when the rest of the panel exceeded 200x, which were the common results of the amplification-based method. Those gaps triggered additional testing. Depending on the panel and intronic coverage requirements, approximately 43% of runs required Sanger fill-ins. The team also designed spike-in probes to improve coverage for many selected targets, which also added significant cost for the assay.

“It was a combination of spike-in probes and Sanger fill-ins, with the latter adding a lot of time to the assay, making it an onerous beast to perform,” Kim says.

The lab was meeting its published four-week turnaround time, but barely. Cases that required reruns or extensive Sanger fill-ins caused additional delay.

Whole-genome sequencing offered a different proposition: one broad assay, less piecemeal testing, and a data foundation that could support multiple needs. Rather than maintaining successive targeted workflows, the lab could sequence broadly, then analyze what was needed: an exome slice, a larger panel, an exome, or eventually the genome.

Kim calls it a way to “future proof” the lab.

Whole-genome sequencing, she says, was “technically the most encompassing platform” and the “simplest workflow.” It also positioned the laboratory to respond as relevant regions of the genome continue to expand beyond traditional coding regions.

Operationally, the move required planning at scale. Michigan Medicine launched clinical whole-genome sequencing alongside an institutional research sequencing effort, working closely with the Michigan Genomics Initiative. That partnership gave the lab volume from banked DNA samples—approximately 100,000 samples in the broader repository, with an initial 10,000-genome pilot—that could be co-sequenced with clinical samples.

That volume mattered. It helped the lab think differently about accessioning, batching, automation, and cost structure. The team had to design processes for bulk accessioning research DNA while individually accessioning and extracting clinical samples, then merge those task lists into a single run list. They built liquid-handling automation for scale and reproducibility and automated data processing, including routing data into separate cloud environments for clinical and research use.

The laboratory also invested heavily in informatics. Existing variant databases had to be migrated into a tertiary analysis environment, and cloud workflows had to be built to support secondary and tertiary analysis without relying on manual initiation of pipeline deployment.

“Envisioning at scale was the key thing,” Kim says, including “bringing in automation, both technical and informatic, wherever we could.”

The efficiency gains were significant. On the technical side, Kim says the whole-genome workflow requires about eight to 10 hours of staff time, including extraction and sequencer loading—less than half the hands-on time of the previous assay. The PCR-free library preparation is also simpler than the earlier exome workflow, and delivers relatively unbiased sequence coverage, which reduces the need for complex workarounds such as spike-in probes and Sanger fill-ins.

That simplicity has consequences. Without amplification bias and with less sensitivity to GC content (stretches of DNA that contain a high proportion of the nucleotides guanine (G) and cytosine (C)), the workflow produces more uniform coverage, Kim says. It also allows the lab to identify important findings that could be missed by an exome-directed panel, including complex structure variations (like translocations, inversions, and copy-number variations) and variants in the non-coding DNA regions. Since launching the assay, for example, pathogenic mutations in the 3’ UTR of a gene were identified that would not have been identified previously.

The early volume numbers are notable. Since going live with whole-genome sequencing last August, the lab has sequenced close to 7000 genomes with Michigan Genomics Initiative (MGI) and is tracking toward an initial target of 10,000 genomes in a year. Turnaround time has also improved. The lab has not yet formally shortened its published four-week turnaround time, but Kim says the team is now meeting it comfortably. Thanks to leveraging a high volume of cases with MGI as well as a foundational partnership with Illumina, the cost of sequencing has been brought down considerably compared to the previous targeted panel. This cost encompasses reagents, consumables, and technologist time.

For staff, the transition brought both relief and uncertainty. When a workflow that once consumed 16 to 20 hours per run suddenly takes eight to 10, people naturally wonder what comes next. Kim says the team addressed that directly: the gained bandwidth would not diminish the staff’s role; it would expand it.

That time is now being redirected toward new validations, new technologies, and higher-value work. Manual pipeline initiation and variant-presentation tasks have been reduced through informatics automation. A growing informatics team—including leadership, informatics faculty liaisons, staff, and project management—now works alongside laboratory teams to support the expanding genomic operation.

The result is a lab that is not simply faster, but more capable. Technologists are less burdened by repetitive, physically intense workflows and more involved in building what comes next. Kim also sees a retention benefit. In a field where technologist turnover can slow operations and drain expertise, better job satisfaction has practical value.

“An added benefit is the decrease in TAT-induced stress for the technologists, although this benefit is harder to quantify,” Kim says, adding that doing “exciting new things” can help improve career trajectories and retention.

For academic medical centers, Michigan’s experience suggests that whole-genome sequencing can be more than a single test. It can be infrastructure: a backbone for consolidating fragmented workflows, scaling operations, retaining send-out testing, and building a more adaptable workforce.

Kim’s advice to other institutions is pragmatic. Understand volume. Build partnerships. Invest in automation. And do not underestimate informatics. A simple, robust liquid handler may be enough for PCR-free whole-genome library preparation, she says, but every instrument has “its own sort of personality and quirks.” The lab jokingly learned it needed a “liquid handler whisperer."

The deeper lesson is that the future of molecular pathology is not just about generating more data. It is about designing systems that can turn that data into durable capacity.

At Michigan Medicine, that meant moving beyond a patchwork of panels, fill-ins, and manual steps toward a single whole-genome foundation. The payoff is showing up in faster workflows, broader reach, reduced complexity, and a laboratory staff with more room to grow—most notably with a stronger foundation and reusable resources to power research.