

Reducing a ketone to a chiral alcohol is one of the most common steps in a synthetic route, and there is almost always more than one way to do it. The question is not which route is more modern. It is which one closes the balance better, on that ketone, at the scale you actually run.
A ketoreductase is worth it when the stereocentre is the problem. The chemical route reaches the alcohol easily enough, but reaching it enantiopure asks for a chiral catalyst, a stoichiometric reagent such as (−)-DIP-Cl, or a downstream resolution that discards half the material. The enzyme puts the selectivity inside the step: it discriminates between the two faces of the carbonyl and delivers the Prelog or anti-Prelog alcohol depending on the enzyme, in water, at 25-35 °C and near-neutral pH. On the atorvastatin intermediate, the one case where the process is published in full, that means over 99.5% enantiomeric excess at 160 g/L substrate loading and 96% isolated yield.
It is worth much less when there is no stereocentre, when the ketone stays insoluble even in a biphasic system and the enzyme will not tolerate the cosolvent it would take, or when the annual volume does not amortise the development. Three honest no’s, and all three are recognisable before you spend.
Four things, in order of weight on the balance.
Selectivity is created in the step, not after it. A classical kinetic resolution has a theoretical ceiling of 50% yield on the enantiomer you want; the rest has to be racemised or discarded. An enzymatic reduction starts from the prochiral ketone: it is a desymmetrisation, and the theoretical ceiling is 100%. That jump shows up first in material cost, and no downstream optimisation recovers it.
Metals leave the process. With no ruthenium or rhodium catalyst there is no transition-metal residue to chase with scavengers or chromatography, and none to justify in the risk assessment under ICH Q3D. On an API that is a chapter of the dossier you never open, not a nuance.
Conditions come down, and the E factor with them. Water or water with a cosolvent, 25-35 °C, atmospheric pressure: no hydrogen under pressure, no cryogenics, no hydrides to quench. It changes the reactor you need and it changes the safety assessment of the step. On the atorvastatin case the reported mass balance is an E factor of 5.8 excluding process water, 18 including it, with butyl acetate recovered at 85%.
You can choose either enantiomer, and the enzyme can be pushed. The initial screening covers enzymes of opposite selectivity: the target decides which one, not the chemistry you happen to have. And where the wild type falls short, directed evolution moves the limit by orders of magnitude: in the atorvastatin process the halohydrin dehalogenase that performs the cyanation was improved more than 2,500-fold over the starting enzyme.
The cost is not the enzyme. It sits in three items that belong in the budget from the start.
Cofactor regeneration. NAD(P)H is never used stoichiometrically: it would cost more than the product. It is regenerated in two ways — a second enzyme with its sacrificial substrate (glucose dehydrogenase with glucose, formate dehydrogenase with formate), or the ketoreductase itself oxidising isopropanol to acetone. The first is more robust and adds an enzyme and a by-product to separate; the second is simpler and requires removing acetone to drive the equilibrium. That choice is a process decision, not a bench one, and it is made by looking at the total turnover number the cofactor has to sustain: that, not the catalogue price of NADPH, decides whether the item weighs. It is the first number we look at when we open a stereoselective reduction.
Substrate loading. An enzyme that performs beautifully at 5 g/L and collapses at 80 has not solved the problem: it has moved the bottleneck to reactor volume. The sustainable loading, the productivity in g/L per hour that follows from it and the substrate-to-catalyst ratio are the numbers on which the two routes are really compared. For scale: the published atorvastatin process runs at 160 g/L with 0.57 g/L of ketoreductase, about 280 grams of ketone per gram of enzyme — a ratio derived from the two reported figures.
Development. Screening, choice of the regeneration system, tuning of the conditions, immobilisation if the enzyme is to be reused. That is time, and it amortises over volume: below a certain annual quantity, an already validated chemical route stays the right answer.
Almost every industrial enzymatic reduction stays inside a dossier. Two do not, and they are the two worth holding on to, because they stand for two different starting points.
Atorvastatin, where biocatalysis rewrites two steps. The intermediate is ethyl (R)-4-cyano-3-hydroxybutyrate. A ketoreductase reduces ethyl 4-chloroacetoacetate to the enantiopure chlorohydrin, with a glucose dehydrogenase regenerating NADPH on glucose; a halohydrin dehalogenase then replaces the chloride with cyanide, in water at neutral pH, instead of a cyanation run under far harsher conditions. The reported figures: 160 g/L loading, over 99.5% enantiomeric excess, 96% isolated yield on the reduction; 140 g/L and 92% on the cyanation; an E factor of 5.8 excluding process water. It is the process that won the Greener Reaction Conditions Award of the Presidential Green Chemistry Challenge in 2006, and it is covered by US 7,807,423 B2.
Montelukast, where it displaces a stoichiometric chiral reagent. Here the ketoreductase reduces a benzoate ketone to the (S)-alcohol, and the title of the process paper says exactly what replaces what: it is an alternative to the asymmetric reduction mediated by (−)-DIP-Cl, a chiral borane used stoichiometrically and at low temperature. Substrate and product are named in full in WO 2009/042984 A1.
The second situation is the one most people who write to us are in: not a process to rewrite, but an expensive chiral reagent to get rid of.
Four questions, in order.
If the first three answers are yes, a feasibility study is worth running. If the fourth says a few tens of kilos a year, it probably is not.
We start from the ketone and the enantiomer you need. A screening of enzymes and microorganisms looks for the right selectivity; the cofactor regeneration system is chosen; the reaction is taken to process conditions, raising the loading to where the enzyme still holds. What you get back is a sheet with the numbers a decision needs: conversion and enantiomeric excess, substrate loading and productivity per litre of reactor, cofactor turnovers and biocatalyst cost per kilo of product, impurity profile against the chemical route.
When the step is one piece of a longer route, the enzymatic reduction is fitted into the route you already have, without rewriting it.
The conclusion, plainly: if your pure enantiomer comes from a resolution today, the comparison almost always turns in the enzyme’s favour, because the 50% you discard is a structural cost and cannot be optimised away. If it comes from a stoichiometric chiral reagent, the montelukast case says it can be removed. If it comes from an asymmetric hydrogenation that is already validated and runs at low catalyst loading, you need volume to amortise the development — and the question shifts from yield to metal residues and to the safety of the step.
The question to start from is not “is biocatalysis right for us”, but “on this reaction, what are the numbers”. If you have a ketone and an enantiomeric excess specification, tell us where you start: the answer is a proposal with those numbers in it.
There is no threshold that holds for everyone: it depends on what the pure enantiomer costs you today. Where the current route is a resolution that discards half the material, the comparison turns early; where an asymmetric hydrogenation already works well, more volume is needed to amortise the development. The honest way to answer is to compare the two costs per kilo of finished product, not per kilo of catalyst.
That is the most frequent case, and it does not close the door. You work with a miscible cosolvent or in a biphasic system, where the organic phase acts as a substrate reservoir and the enzyme stays in water. The real constraint is how much solvent the enzyme tolerates before it loses activity: that is measured in the screening, not assumed.
Yes, if it is immobilised on a support, and then the biocatalyst cost divides by the number of cycles it withstands. It pays when the enzyme is the heavy item in the balance and the process is repetitive; on a single campaign the extra step often is not worth it. We assess it together with the choice of the regeneration system.
Technical content and news from our laboratories, sent when there is something worth saying.