Science1 publisher2 min readPublished
Nirrin asks bioprocess labs to trade chemometric models for a library of known spectra
Nirrin's chief executive says its tunable-laser near-infrared platform reports concentrations directly, so one method built in process development runs unchanged in manufacturing, and five to 10 large drugmakers already use it in development.
The Scientist · Science desk
What happened
- Nirrin Technologies CEO Bryan Hassell told GEN that empirical spectroscopy models are built from large designed experiments, are tied to the process and instrument they came from, and must be rebuilt when process, scale or instrument changes.
- He put pure-component analysis forward as the alternative: each component has a known spectrum, the measured spectrum is resolved into concentrations directly, and the method skips the design-of-experiments campaign.
- The company's tunable laser near-infrared platform, which uses a laser in place of a broadband lamp, will be shown at the Bioprocess International East conference in Boston.
- Hassell said a TL-NIR protein measurement is five times faster than variable-pathlength UV and needs no dilution, no sample preparation and no pathlength adjustment.
- Between five and 10 top biopharmaceutical companies currently run TL-NIR systems in process development, according to Hassell. He named none of them.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- decision An analytics group weighing this now is deciding on one company's account of its own instrument, because the cross-site dataset behind the transfer claim that carries the business case remains unpublished.
- constraint The approach is bounded by how completely the spectral library describes the stream. A missing absorber throws no error and moves the concentration the engineer acts on.
- cost If the savings hold, they come out of validation campaigns and outsourced assay invoices, so the purchase has to be argued by analytical development and QA rather than won on instrument performance.
- capability A reading that arrives as a concentration in seconds or continuously can feed process control directly, where a spectrum needing model interpretation cannot.
Hassell's description of pure-component analysis contains a condition that the words "model-free" cover over. A fit that resolves a measured spectrum into the concentrations of components with known spectra can only assign absorbance to components in that set [5]. Anything else in the stream that absorbs still absorbs. Its signal has to go somewhere, and where it goes is the concentrations being reported [15]. Empirical chemometric models deal with that background by learning it, which is what designed experiments spanning every condition the method will meet are for [19].
The pure-component approach is fitted to no particular vessel or instrument, and the transfer argument rests on that. "A method built in development runs the same way in manufacturing, and on a different system at a different site," Hassell said [6].
The tunable laser is a second claim, and it stands apart from the first. Putting more optical power into each wavelength lowers the noise on the absorbance measurement [8]. The set of components the fit knows about stays the same [18].
Spectroscopy on a process stream exists to give protein and analyte numbers in real time [1], and the savings Hassell describes arrive in two different units. One is seconds per measurement, against a benchtop protein assay [10]. The other is days, the turnaround on samples sent out for HPLC, along with the cost of each sample [11]. A site that already runs titer in-house collects only the first of those.
Every figure in the GEN account is Hassell's, and the article does not include a cross-site transfer study or a customer name [16]. The count he gave is also soft at both ends: its upper bound is twice its lower [17]. On the next step he was plain. "We are at the beginning of bringing this into manufacturing environments," he said [13], adding that the company is "now working with key customers on manufacturing deployment from there" [14].
Transferability is testable. Run one pure-component method with one spectral library on two instruments at two sites across a change of scale, and compare the reported concentrations against an independent reference assay. That comparison would show whether the method transfers. Hassell's own order of operations puts process development first, where he said the technology is proven, validated and transferability demonstrated [14].
What to watch
- Publication of a cross-site, cross-instrument comparison using one pure-component library, checked against an independent reference assay.
- Whether any of the process development users goes on record with a commercial production deployment, and in which regulatory filing it appears.
- Data on how the pure-component method behaves in media with uncharacterised absorbers, such as high cell density or antifoam additions.