Reinforcing the Supply Chain of COVID-19 Therapeutics with Expert-Coded Retrosynthetic Software

Abstract

Supply chains become stressed when demand for essential products increases rapidly in times of crisis. This year, the scourge of coronavirus highlighted the fragility of diverse supply chains, affecting the world’s pipeline of hand sanitizer, 1 toilet paper,2 and pharmaceutical starting materials. 3 Many drug repurposing studies are now underway. 4 If a winning therapeutic emerges, it is unlikely that the existing inventory of the medicine, or even the chemical raw materials needed to synthesize it,5 will be available in the quantities required to satisfy global demand. We show the use of a retrosynthetic artificial intelligence (AI) 6-10 to navigate multiple parallel synthetic sequences, and arrive at plausible alternate reagent supply chains for twelve investigational COVID-19 therapeutics. In many instances, the AI utilizes C–H functionalization logic, 11-13 and we have experimentally validated several syntheses, including a route to the antiviral umifenovir that requires functionalization of six C–H bonds. This general solution to chemical supply chain reinforcement will be useful during global disruptions, such as during a pandemic.

A computer algorithm to discover iterative sequences of organic reactions

“…To begin with, the computer is provided with appropriately coded rules that describe the individual reactions, here taken mostly from Chematica's collection [15][16][17][18] (>100,000 rules) or, in some analyses, from that of Allchemy (~10,000) 19 . These expert-coded rules were chosen owing to their high quality 20 , which has been documented by numerous experimental validations of computer-designed syntheses, including those of drugs 16,21 , prebiotic compounds 19 and complex natural products 18 -this quality is essential for the correctness of the reaction sequences we set out to discover. Briefly, each reaction rule is based on the underlying reaction mechanism (more specifically, its substituent scope is broader than any specific literature precedent), it considers around 400 potentially incompatible functional groups as well as groups that require protection, and provides typical reaction conditions (for screenshots see Supplementary Figs.…”

Chemist Ex Machina: Advanced Synthesis Planning by Computers

“…Figure e, taken from our recent work, shows some patent-circumventing routes Chematica found for a blockbuster drug called sitagliptin. Avoidance of specific reactions and hard-to-source key substrates is yet another option, which we and others have shown useful in seeking alternative paths to important antiviral compounds (of note, in ref , Cernak’s group validated such plans by experiment). Finally, Figure illustrates a situation in which there are multiple and isotopically labeled targets (M+6, 13 C isotopically labeled derivatives of ten anticoagulant rodenticides), and the program is asked to develop a global synthetic plan to make all of them in the most economical way, hopefully sharing between individual routes as many intermediates as possible and using an economical source of 13 C atoms.…”

PhactorTM – a High Throughput Experimentation Management System

“…In a second example, an oxidative indolization reaction screen, inspired by the conditions of Glorius, 11 was performed using copper catalysts and ligand/additive combinations to optimize the penultimate step towards a synthesis of umifenovir via the coupling of 4 and 5 to produce 6. 12 distributed into the four rows while combinations of magnesium sulfate (0.0 equiv or 1.0 equiv) with 2-(1H-tetrazol-1yl)acetic acid (L1), or 2,6dimethylanilino(oxo)acetic acid (L2) at 40 mol% were distributed into the columns as DMSO solutions, with 3.0 equiv. of cesium carbonate added to every well as a suspension in DMSO.…”