Solving a Problem AI Made Worse: Axor Biosystems and the Sample Prep Squeeze
- washburnadam
- 2 minutes ago
- 6 min read
Imagine a busy airport where the runways have been expanded to handle ten times more flights, but the security lines, baggage handling, and gate agents are all stuck at their original capacity. Planes are able to take off faster than ever, but passengers are piling up at the gates. That, in essence, is what has happened to modern biology.
Sequencing throughput and AI-driven analysis have raced ahead, while the slow, manual process of preparing biological samples to feed those systems has not kept pace. Labs can generate terabytes of data, but only if they can prepare enough samples to keep the runway full.
Recently, AnalyticalChemistryStartups spoke with Abizar Lakdawalla, CEO and co-founder of Axor Biosystems, about how his team is rethinking sample preparation from the ground up — and why the answer turned out to be hidden in a piece of lab equipment that has been around for decades.
The Founding
Lakdawalla's career has been spent close to the engines of the genomics revolution. He was an innovative researcher at Illumina during the years when the cost and accessibility of sequencing evolved dramatically. He went on to Ultima Genomics, another high-throughput sequencing company, where the data-generation problem only grew larger. The progression, as he describes it, was almost mechanical: as sequencing throughput climbed, the question shifted from "how do we generate this data?" to "how on earth do we analyze terabytes of it?" That second question found its answers in machine learning. However, that began the process of moving the bottleneck upstream.
By the time the data analysis problem had eased, Lakdawalla and his colleagues found themselves facing another, less glamorous one. Sample preparation — the indispensable work of cleaning up reactions, purifying biomolecules, and quality-checking the results — had become the new rate-limiting step.
In addition, funding cycles shifted with political winds, and samples were being stockpiled in freezers in the hopes that money would return. Meanwhile, no one had figured out how to process hundreds of thousands of samples at low cost. When Apton Biosystems was acquired by PacBio, Lakdawalla and Brett Anderson, who had previously been colleagues at Apton Biosystems, started looking for the next challenge.
The two had hit it off during their two or three years together at Apton. Lakdawalla had moved from the Bay Area to the Seattle area in what he describes, with some amusement, as the early stages of an anticipated retirement. That retirement did not take. As he told ACStartups, "you get addicted to startups." The pair began exploring ideas, reading widely, and consulting with a small group of advisors. Industry feedback was encouraging, and Axor Biosystems was founded in January 2024.
The Problem: A Bottleneck AI Made Worse
The framing Lakdawalla returns to repeatedly is that Axor is trying to solve a problem that AI has made worse. As compute-side and analysis-side bottlenecks dissolved, demand for processed biological samples climbed sharply. Axor's own materials estimate that AI-driven biology requires roughly ten to a hundred times more processed samples to deliver on its promised breakthroughs in diagnostics, therapeutics, and industrial biosciences. The catch is that the workflow necessary to produce those samples has not been rebuilt for that scale.
Today's standard sample preparation pipeline involves stitching together separate instruments and protocols: gel electrophoresis for analysis, column- or bead-based purification, robotic liquid handling for parallelization, separate quality-control steps, and so on. Each handoff adds time, cost, error, and capital expenditure. The current standard in next-generation sequencing library cleanup, for example, is bead-based purification run on liquid-handling robots — a workflow that works but does not scale gracefully into the hundreds of thousands of samples per year that newer applications demand.
The Technology: Old Physics, New Geometry
What Axor has built is, by Lakdawalla's own description, not a new invention so much as an unfamiliar combination. The principle is electrophoresis — the movement of charged molecules through a matrix under an electric field. This technology has been a workhorse of molecular biology for decades. Gel electrophoresis separates molecules by size. Preparative electrophoresis goes a step further and lets researchers physically recover those separated fractions. However, running it on the standard slab-gel format is inherently low-throughput.
The Axor insight was geometric. By moving from a planar slab into a three-dimensional structure that fits within the footprint of a standard microplate well, the team was able to preserve the resolution of conventional electrophoresis while gaining a volumetric advantage. Axor's consumable is twelve independent strips of eight cells each, with every nine-millimeter-square cell functioning as its own self-contained electrophoresis unit. The result is ninety-six fully independent electrophoresis processes running in parallel, each one electrically isolated and individually controlled.
That independence is what unlocks the rest of the system. A printed circuit board sits on top of the consumable and routes current to electrodes in each cell. Fluorescence imaging monitors all ninety-six cells in real time, and a feedback loop adjusts the current in each one to push molecules faster or slower, or into different channels. Because the system is doing electrophoresis, two pieces of information come essentially for free: molecular weight, from the separation pattern, and quantitation, from fluorescence intensity and area-under-the-curve calculations. Size selection (for example, pulling out a five-hundred-base-pair DNA fragment within a defined window) happens electrically rather than through a series of bead binding and wash steps. Pumps, syringes, magnetic separators, and the robots that orchestrate them all collapse into one box.
Lakdawalla is careful to ground these claims. He noted to ACStartups that the Axor team wants to innovate without overstating impact. But the comparative numbers are striking on their own terms: roughly ten times faster than the fastest robotic processes for equivalent workflows, with what Axor describes on its website as a path toward an overall hundred-times benefit when capital, consumables, and hands-on time are tallied together.
At a hundred thousand samples per year, Axor estimates customer savings of up to $4.5 million annually. The user experience is straightforward: pipette samples into the input wells of a strip using standard 384-well spacing, start the run, and come back fifteen to twenty-five minutes later to pipette out purified product from the corresponding output wells. Comparable lab-robot workflows can take most of a day.
The Beachhead and What Comes Next
Axor's initial commercial focus is sequencing library cleanup at genome centers — the large institutions with dedicated technology evaluation teams that vet new platforms before pulling them into production. It is a deliberate choice. The bead-based cleanup standard is well understood, the pain points are well documented, and the throughput math works in Axor's favor. The company has also been showing the system at conferences like the Society for Laboratory Automation and Screening, where Lakdawalla said the early interactions have been productive.
From voice-of-customer work, the next unmet need that surfaced was on the protein side. Customers wanted the ability to run, for example, an isothermal step that converts DNA to protein. And then be able to separate, purify, and quantify the product within the same consumable.
Axor's microplate format also supports standard SDS-PAGE-style workflows. Because the imaging hardware functions as a fluorescence plate reader, there are other applications the team can consider. Other applications on the horizon include separating viruses and using the system as an affinity-electrophoresis platform, where immobilized DNA probes or antibodies in the gel matrix can capture biomarkers directly from plasma and produce a multiplexed readout in a single cell.
Challenges: Scaling the Manufacturing, Not the Idea
When ACStartups inquired about the biggest challenges ahead, Lakdawalla noted manufacturing scale-up first. The instrument itself is compact — it is roughly a seven-inch cube. However, each consumable strip carries an integrated PCB, and getting that supply chain to commercial volume is a complex engineering and operations feat.
A second challenge is recruiting. Finding individuals with the right mix of skills is never an easy proposition. Focus is the third quiet challenge. The platform's breadth — analytical, preparative, and affinity electrophoresis across DNA, RNA, protein, viral, and cellular samples — is a strength, but also a temptation. Staying focused on their main goal and key customers is the key to generating a genuinely useful product.
Clearing the Gates
Lakdawalla is candid about what Axor is and is not. The underlying physics is well known. Electrophoresis is decades old. Microplates are decades old. What is new is the geometry, the matrix and buffer chemistry that preserve resolution in a miniaturized format, and the per-cell electrical control that makes ninety-six independent runs possible in parallel. Everything Axor is doing is known, but now combined in an innovative way.
Which is, in a sense, exactly what the example airport from the introductory paragraph needed. The runways are still the runways. The planes are still the planes. What was missing was a way to move passengers through the gates fast enough to match the pace of everything downstream.
If Axor's microplate scales the way the team believes it can, the bottleneck that has been quietly holding biology back — the slow, expensive, error-prone work of getting samples ready — starts to look less like a permanent feature of the field and more like an artifact of the equipment we happened to inherit. Ten to a hundred times more samples processed for the same money is not a small improvement. For gene therapy, cancer therapeutics, and any field whose progress depends on volume as much as ideas, it is the difference between flights that take off and passengers who never make it past security.
To learn more about Axor Biosystems and its preparative-electrophoresis-in-a-microplate platform, visit axorbio.com.