The Empty Lane Problem
by Gabor Bethlendy
The ‘Empty Lane Problem’ is an understated issue in the Next-Generation Sequencing (NGS) space with a big cost/sample impact. It turns out that NGS economics is very sensitive to capacity utilization. The same principle that has rapidly reduced the cost of sequencing – parallel processing of samples – works against you if you don’t fully load a run.
At Meenta, we think of lanes on sequencers like seats on an airplane.
Financial sensitivity to capacity utilization is found in almost every industry. In the airline industry, for example, with its high fixed costs, airlines want planes to operate with as many passengers at as high a price as possible. But there is a trade off between demand and price (ignoring a few other considerations such as schedule). Fly a route too often and the airline can’t profitably fill enough seats.
Airlines manage this problem in different ways depending on the options available to passengers. On popular routes served by multiple airlines passengers can shop for prices, departure times and seating options of their choice. On these routes airlines deploy logisticians with fancy computer algorithms to optimize flight frequency vs. filled seats at various prices to make up their schedules. The optimum mix is probably some complex version of a Marginal Revenue = Marginal Cost graph from Economics 101 with some factor figured in for profit tolerance. On captive routes serviced by only one or few airlines passengers are at the mercy of the airline schedule with minimal options for time or price. On these routes airlines make sure they have enough passengers at a desired price point by creating an inventory of passengers to fill seats. They make you wait. The trick on these routes for airlines is not to create so much excess demand as to entice competition.
Services like Priceline and Orbitz have built their business models around these problems: discounting pricing to fill unsold seats with algorithms to adjust pricing in real time.
Sequencing labs operate like a captive airline route: they wait to run a sequencer until they have samples for all the lanes. Every core lab is different, but wait times can vary from weeks to months. Price/cost economics is more important that the researchers/clinicians inconvenience or time. The obvious difference between airplanes and sequencers is that airlines can move their equipment (airplanes) to meet demand where sequencers are generally stuck in a location - bit paradoxical that the gazillion pound machine moves and the almost portable machine doesn’t...
The solution is to provide a live marketplace for core labs to fill unused lanes with samples.
Both users and sequencing facilities benefit. With live, lane-level access to sequencers, Researchers and funding providers can now factor the individual price of lane into their grant and experiment budgets. For researchers, the experiment design and budgeting can focus on short term changes in their experiments, and a per lane price. Researchers can break experiments down, and process samples across multiple trusted core labs.
For Core Labs and large research institutions, the process of filling lanes to maximize instrument utilization becomes much easier, decreasing queues and wait times. Cores can factor higher utilization numbers into their pricing.
Everyone benefits.
This is the goal: change the underlying assumptions and economics of equipment ownership. The result is faster and cheaper science.
The Complexity Problem: Matching Samples To Sequencing Lanes
Running an Illumina HiSeq with 2 of its 8 lanes empty is more impractical for a core lab than it is for an airline company. For science and research this problem is described as the ‘Empty Lane Problem’, where the passengers are the samples or lanes on a sequencer. Unlike the travel industry, which can afford to build in some unsold seats on a flight, equipment owners are much more restricted in their options.
Equipment owners have few techniques for adjusting to unsold or missing lanes. In practice core labs solve this issue by building up a de-facto sample queue. They assume their instrument will have downtime because it can not be run half full, so factor the cost of downtime in their per sample price. And when demand is too variable, the core lab simply do not buy more equipment.
While pricing and equipment utilization is important in the grand scheme, the impact of missing lanes impacts researchers and their project on a daily basis. When a researcher comes to a lab with samples they need to run on a given instrument, the core lab manager needs to fit their request into the core labs queue of samples, both on a first come/first serve basis, but also on a equipment and configuration basis. The question is often, do I have any samples in the queue that can be combined with what this researcher needs. If the answer is yes, then core lab tells both the waiting sample owner and the new sample owner when the sample will run. If the answer is no, then the core lab puts the sample in its queue, waits for a match, and tells both researchers to wait.
The more variations of lanes and configurations in the queue, the harder it is to fill lanes and schedule the sample to run. The impact of this issue is that samples wait in equipment queues, researchers wait for data and equipment stands unused. Waiting drives up costs and slows down research. While research equipment was never purchased or priced with any assumption of 100% utilization, the impact of missing lanes impacts every lab and research facility.
Here is a Tweet-able summary… “How The Empty Lane Problem Increases The Cost Of Next-generation Sequencing.”