Counting steps

Abstract

The use of step count as a metric of synthetic efficiency carries opportunities and challenges. Here, proposals are made to standardize what constitutes a synthetic step and how steps are counted. These proposals may be beneficial in the holistic evaluation of published synthetic routes.

Early in the movie Interstellar, the engineer and pilot Cooper is engaged in a difficult conversation with his son Tom’s school principal. The principal has just informed Cooper that, on the basis of recent standardized test results, Tom has been relegated off the college track¹:

Principal: Tom’s score simply isn’t high enough.

Cooper: (pause) What’s your waistline? 32? Maybe 33 inseam?

Principal: I’m not sure I see what you’re getting at.

Cooper: You’re telling me it takes two numbers to measure your own ass but only one to measure my son’s future?

A reader of the synthesis literature may empathize with Cooper’s consternation when confronted with the task of evaluating syntheses that are metricized by a single number: step count. As scientists, we value hard data. To push the science of synthesis forward requires an understanding of what data we need and what we are trying to improve. In considering how many steps are required to convert starting material A to target B, two interrelated but distinct questions immediately emerge: first, what constitutes a chemical step? And second, where does step count fit in as a metric for the quality of a synthesis?

Defining a synthetic step

Presupposing for a moment that knowing the step count of a synthesis is valuable for a reader and/or potential adopter, it follows that any value derives from a common understanding of what defines a step. A casual observer of our modern world will immediately grasp that misunderstandings arise and different conclusions are reached when conflicting definitions are applied. Simple can often be best, and it is appealing to gravitate to a definition that was explicitly put forth in a recent perspective²: “…a step is defined as an operation that does not involve any intervening purification/separation, including removal of solvent…”. The components and usefulness of this definition can be parsed at several levels.

First, it is helpful to consider how and why steps are counted for mechanistically distinct reactions that occur in the same pot. A recent case that has been discussed at length is the protection of an alcohol as a silyl ether followed by allylic oxidation using selenium dioxide (SeO₂)^3,4. Operationally, this sequence is carried out by waiting for the completion of the protection, followed by direct addition of the oxidizing reagent, SeO₂. While mechanistically decoupled, the operations occurred in the same pot and were counted as a single step. The telescoping of reactions — the carrying out of distinct transformations in the same pot without purification or solvent changes — combines steps, and has a significant impact on operational efficiency. The chemist who is telescoping reactions, like the chemist who strives for the optimal synthesis strategy⁵, is thinking proactively about both maximizing efficiency and how their method could translate for other users in the most positive way possible. Since solvent use is a major driver of synthetic inefficiencies (as measured by process mass intensity, PMI⁶), pot economy^7,8 constitutes a way to achieve a synthesis in a more sustainable manner. In any evaluative metric (such as step count), the achievement of this aspirational goal should be encouraged. As a result, step count can be considered as a crude reflection of synthetic efficiency that requires the most valuable of resources — time. Therefore, the lower step count that arises from telescoped reactions provides useful information for the reader.

Consider the hypothetical synthesis of the aldol adduct 5 by two different routes (Fig. 1a). The first route relies on the sequential stoichio-metric generation of metalloenolate 2 through reaction of ketone 1 with a strong base, introduction of an aldehyde electrophile (3, which must be done separately to avoid undesired reactions) to achieve C–C bond formation, and liberation of the neutral, isolable product 5 by treatment of the metal alkoxide 4 with acid. The second route makes the same compound but uses the Mukaiyama variant wherein a neutral enol silane 6 is prepared, then treated with the aldehyde and a catalyst, resulting in liberation of the final product by cleaving the silyl ether 7 with aqueous acid. Most chemists would assign a single step to the first route, even though a sequence of three mechanistically unrelated reactions must be orchestrated in a prescribed order to achieve the overall synthesis. While the enol silane and secondary silyl ether of the second route are, in principle, isolable, in this hypothetical example they are processed in the same pot without purification. The parallel nature of the routes implies that the second route is also one step, as would any catalytic direct aldol reaction that forms alcohol 5 from ketone 1 and aldol 3.

Fig. 1 |. Defining and reporting synthetic steps.

a, A comparison of enol/enolate bond constructions for aldol adduct 5. b, A single-step three-component, one-pot coupling in the synthesis of X-206. The single step⁹ involves: i) 8 (1.00 equiv), ^tBuLi (1.93 equiv.), Et₂O, −78 °C; ii) 9 (0.86 equiv.), −78 °C to 0 °C; iii) LDA (1.5 equiv.), 0 °C; iv) 11 (0.67 equiv.), 0 °C. c, What can we do as synthetic chemists? A common understanding of what constitutes a synthetic step combined with standardized reporting of select metrics can enhance the value of these data to readers of the chemical literature. ⁱPr, iso-propyl; LDA, lithium diisopropylamide; MOP, methoxyisopropyl mixed ketal; ^tBu, tert-butyl; Tf, trifluoromethanesulfonyl; TMS, trimethylsilyl.

A more complex example underscores this point. The convergent synthesis of the polyether antibiotic X-206 (Fig. 1b) hinges on a three-component coupling step using a conjunctive bifunctional fragment that sequentially acts as an electrophile and nucleophile in the same pot⁹. The organolithium derived from the vinyl bromide 8 reacts chemoselectively with the Weinreb amide 9 to give the tetrahedral intermediate 10. From that point, metalation of the dimethylhydrazone enables coupling with epoxide 11 that delivers (after workup) the complex product 12. As stated by the authors⁹, “…the effectiveness of this route stems from the consolidation of the assemblage process into a single reaction.” The illustrated one-step coupling was possible, whereas separating the constituent operations would have had low probability of success: the chelated lithium alkoxide 10 provided in situ protection for the latent enone functionality that would have been incompatible with the ensuing step.

A standardized approach (Fig. 1c) to define step count may assume that the following conditions apply:

• Telescoped reactions, of the type illustrated above, which occur in one pot and with no change in solvent, should be counted as one step

• A solvent switch, involving either partial or complete evaporation and the introduction of a new solvent, marks the end of a step

• A filtration of any sort (such as the use of a frit, filter paper or silica plug) marks the end of a step

• Workup, isolation, and/or, if necessary, purification marks the end of a step

The value of reporting step count

If it is possible to have a common framework for categorizing synthetic steps, a broader issue that emerges is how step count is valued as a metric. Cooper’s discussion with the principal is germane here, as we can recognize the advantage of evaluating synthetic work through a lens broader than just the step count. Indeed, an adjective that appears frequently and appropriately in the treatment of this topic is ‘holistic’. Delving more deeply, to evaluate a synthesis, we have an array of approaches and criteria available to us, including, but not limited to:

• Material and operating costs⁶

• Conversion cost factors⁶, which encompass criteria including atom economy¹⁰, yield, volume-time output, environmental factor (E-factor, the ratio of mass of waste to mass of product), or process mass intensity; there are also merged or aggregated variants that measure these key performance indicators¹¹

• The use of a radial pentagon to report various material efficiencies¹²

• Pot economy^7,8

• Strategic efficiency index: step count versus molecular complexity⁵

Full evaluation of these analyses is beyond the scope of this Comment; rather, it must suffice here to convey that step count provides only one piece of data that, while valuable, should be deployed in a holistic manner with other measures of synthetic efficiency.

A recognition can be made that the goals of different syntheses can be quite distinct. At the limit, a discovery project would seek divergent access to large numbers (and potentially small quantities) of compounds, while a process project requires delivery of large quantities of a single compound in a sustainable manner¹³. Some of the metrics that would be most appropriate for the latter may be ill-fitting for the former, which bears some similarities to a synthesis that seeks to verify the structure of a bioactive natural product available in small quantities. In an academic setting, syntheses that offer key information about reactivity, strategy, and design tend to be the most impactful and subsequently advance the field in general. The heightened interest in transparency and rigour is a positive for the field and chemists could err on the side of inclusion of some of these metrics in the self-evaluation of their work within a given manuscript. Minimally, the regular documentation of the following information in the conclusion section of synthesis papers could be useful to the reader (Fig. 1c):

• The total and longest linear sequence step count, using the above definitions

• The starting material, as executed in practice, should be defined as well as its cost and source at the time of purchase (or a current cost) for quantities required

• The lowest overall yield from the defined starting material. In most cases, this yield would be calculated from the longest linear sequence, although scenarios are envisioned where the shorter path of a convergent sequence could have a lower overall yield Step count features explicitly in two of these three metrics, but is contextualized for the reader in a deeper manner than a single number. The field may collectively benefit when like-for-like comparisons are made through the lens of identically-used and universally understood terms.