

People who bring us a kinetic resolution usually bring it in order to have it taken away. The sentence it arrives with is almost always the same: “we throw away half the material, there must be a better way”.
There almost always is a better way. It is almost never the one people expect.
The 50% ceiling of a kinetic resolution is real: the enzyme converts one enantiomer and leaves the other, and on a racemate half the material never becomes product. But that ceiling, on its own, says almost nothing about cost.
What decides cost is where in the sequence that 50% is paid, and immediately after that, what the selectivity you actually have is worth. Those are two different calculations, and they are done in that order.
The selectivity of a kinetic resolution is expressed as the enantiomeric ratio E: the ratio of the specificity constants kcat/KM of the two enantiomers. It is not a percentage and it is not an enantiomeric excess. It is a property of the enzyme-substrate pair under those conditions, and it stays the same as the reaction proceeds.
What changes as the reaction proceeds is the pair conversion and enantiomeric excess, and it changes in opposite directions for product and remaining substrate. The relationship is quantitative and has been known since 1982, in the equations of Chen, Fujimoto, Girdaukas and Sih, which for an irreversible resolution link E, conversion c and the enantiomeric excess of the product.
From there follows a counter-intuitive and very practical fact: at fixed selectivity, the enantiomeric excess of the product falls as conversion rises. With E = 200 the product comes out at 98.0% enantiomeric excess at 40% conversion, 97.6% at 45% and 97.0% at 48%. The remaining substrate runs the other way and is enriched: at 45% conversion it sits at 80%, at 48% at 89.5%.
This means that a selectivity figure without the conversion at which it was measured says nothing, and that an enantiopurity specification is never read on its own. If you need 98% on the product and you have E = 200, you have to stop at 40.6% conversion: below the theoretical 50% ceiling there is a second, lower one, and it is your specification.
The first industrial route to pregabalin resolved the racemate at the last step, with (S)-mandelic acid, and the unwanted enantiomer was hard to recycle. Overall yield: 25-29%.
The route that replaced it did not make the resolution dynamic. It moved it to the first step of the process, onto the racemic cyanodiester that comes in as raw material, with a Thermomyces lanuginosus lipase (EC 3.1.1.3, commercial preparation Lipolase) selected in screening for an enantiomeric ratio above 200. The reaction reports over 98% enantiomeric excess at 45% conversion, at 765 g/L on total volume and with a protein loading around 0.8%. The diester of the other configuration is racemised and returned to the cycle.
It is worth pausing on those first two numbers, because taken together they say more than they declare.
First of all, that 98% necessarily refers to the product. Mass balance requires the enantiomeric excess of the remaining substrate to be that of the product multiplied by c/(1−c): at 45% conversion the residue cannot exceed 81.8% even with a perfectly enantiopure product. There is no ambiguity about which of the two materials sits at 98%.
With that settled, the numbers close. An E of exactly 200, at 45% conversion, would give a product at 97.6%: below specification. To stay above 98% at that conversion you need E ≥ 245. The screening’s “greater than 200” is therefore a lower bound — typically the ceiling of the assay — and the enzyme taken into production is appreciably more selective than that. It is the kind of reading worth doing on every screening report that reaches you.
The process result, with the theoretical ceiling on the single step unchanged:
The eight-fold on solvents and the 765 g/L are the same number seen from two sides. A resolution running at three quarters of a kilogram of substrate per litre does not need the solvent volume of one running dilute, and in a pharmaceutical mass balance solvent is almost always the dominant line: in the old route it is 86% of inputs.
The enzymatic step was taken to 10,000 kg and production switched to the new route in the third quarter of 2006. On an annual basis the difference is worth 800 tonnes of starting material, 500 tonnes of mandelic acid and several million gallons of solvent.
It is worth reading the first line of that list next to the third. Overall yield improved by more than ten points; material consumption fell nearly five times. They are not the same improvement, and the second does not come from the selectivity of the enzyme: it comes from where the resolution was placed and from the concentration at which it runs.
This is the point the pregabalin case makes concrete. A resolution placed at the end of a synthesis discards material on which every preceding step has already been spent; the same resolution placed at the entrance discards raw material. The chemistry is identical, the balance sheet is not, and the difference is not marginal: it is the jump from an E-factor of 86 to one of 17.
This is why the first question we ask in front of a resolution is not “can it be made dynamic?”, but “where is it?”
There are three ways past 50%, and they cost differently.
External recycle. The remaining enantiomer is isolated, racemised in a separate operation and returned to the cycle. This is what the pregabalin process does. It costs one unit operation and a fraction of yield on every pass, it introduces no new catalyst into the reactor, and it stops when recovery no longer pays. It is the boring route, and it is almost always the first one to evaluate.
In situ racemisation. The slow enantiomer is reconverted while the enzyme works, and the theoretical yield returns to 100%: this is dynamic kinetic resolution. It requires the racemisation catalyst and the enzyme to coexist — same temperature, same solvent, neither deactivating the other — and it is that compatibility, not selectivity, that is the design problem.
Desymmetrisation. You start from a prochiral or meso substrate rather than from a racemate. Here the 50% ceiling does not exist, because there is no enantiomer to discard, and the behaviour of the system changes too: the enantiomeric excess does not decay with conversion, because there is no second enantiomer accumulating. Where the sequence allows you to reach a meso or a prochiral intermediate, it is the cleanest of the three. These are the three routes we work on in kinetic resolutions and desymmetrisations.
It is worth looking at real numbers rather than at a scheme.
On the dynamic kinetic resolution of 1-phenylethanol, with immobilised Candida antarctica lipase B (0.5 mg/mmol) and a ruthenium catalyst for in situ redox racemisation, a one-mole run at 70 °C for 20 hours, with 0.05 mol% ruthenium and isopropenyl acetate as acyl donor, gives 159 g of product, 97% isolated yield and over 99.8% enantiomeric excess. The 50% ceiling is broken, and broken well.
The same work reports the next scale up, and that is where something is learned. At ten moles, with ruthenium lowered to 0.01 mol%, it took twenty-one days to reach 87% yield and 97% enantiomeric excess. And the detail that explains it: the unreacted alcohol was left at 19% enantiomeric excess. If racemisation kept pace with the enzyme, the remaining substrate would tend towards the racemate, that is towards zero: that 19% says the metal could not keep up at that loading. The catalyst is also oxygen sensitive: under an oxygen atmosphere the yield falls to 60% in sixteen hours, against 98% in three hours under argon.
The conclusion is not against dynamic resolution: it shows where the constraint sits. In a dynamic resolution the bottleneck is almost never the enzyme. It is the racemisation rate at the metal loading you can afford, and those are two things that can only be seen together.
There is then a regulatory consequence to account for before starting rather than after. Ruthenium is a Class 2B element under ICH Q3D(R2), with a permitted daily exposure of 100 µg/day oral, 10 µg/day parenteral and 1 µg/day inhalation. Class 2B elements can normally be excluded from the risk assessment precisely because they are unlikely to be present, unless they are intentionally added. A ruthenium racemisation catalyst adds them intentionally: the risk assessment opens, and with it an assay method, a specification and a removal strategy. On an intermediate far from the active substance this is often not a problem; on the last step it can be a large one.
Four questions, in this order.
If the answers are “not much, low, yes, no need”, the kinetic resolution you already have is probably the right answer, and the work to do is on selectivity and enzyme reuse, not on architecture.
On a resolution we measure the quantities that enter cost per kilogram: the enantiomeric ratio as a ratio of specificity constants, the pair conversion and enantiomeric excess on both product and remaining substrate across the whole reaction and not at a single point, the effect of solvent, acyl donor and temperature, and how the immobilised enzyme holds up over cycles.
On that last point the number that counts is not initial activity: it is the total turnover number and the kilograms of product per kilogram of biocatalyst, that is how much work the enzyme does before it is replaced. Initial activity photographs well and predicts little; the residual activity curve over cycles is what enters the economics.
Where the sequence allows it we also look at desymmetrisation, for the reason given above: it removes the problem instead of working around it.
If you have a kinetic resolution in production or in development, the useful conversation takes twenty minutes and starts from two things you already have: where it sits in the sequence, and at what conversion you are reading the enantiomeric excess. In two cases out of three you can tell before opening the laboratory whether there is a large gain or only a small one.
Kinetic resolutions and desymmetrisations are one of the routes we work on in biocatalytic development. From there you can also see the rest of the section.
Because the enzyme converts one enantiomer of the racemate and leaves the other: on a racemate, half the material never becomes product. The limit applies to the step, not to the process: by recycling the discarded enantiomer, by racemising in situ or by starting from a prochiral substrate, the process goes past it.
It depends on conversion, and the dependence is strong. For an irreversible resolution, with E = 200 the product sits at 98% only by stopping around 40% conversion; at 45% you need E ≥ 245. A selectivity quoted without the conversion at which it was measured is not a specification.
No, and this is the most common mistake. The larger gain usually comes not from in situ racemisation but from moving the resolution to where the discarded material is cheap. In the published pregabalin process the resolution stayed kinetic and moved from the last step to the first: overall yield from 25-29% to 40-45%, E-factor from 86 to 17. It should be added that a metal racemisation catalyst brings that metal into the elemental impurity assessment under ICH Q3D.
All the figures quoted come from published literature, guidelines and public documents: no project of ours.
Technical content and news from our laboratories, sent when there is something worth saying.