Lee Cronin, a professor at the University of Glasgow, explains the Chemputer – a universally programmable device for synthesizing any molecule.  Development is ongoing and proof of concept has been achieved for several molecules.  He estimates that with the right components it is capable of performing 95% of all organic chemical synthesis.


Presentation: The first programmable Turing complete chemical computer

  • Robotic control of chemistry is a standard feat nowadays, but standardization of robotic chemistry remains out of reach.  The specialized machines used can perform one or two specific feats repeatedly.  Lee Cronin is on a quest to create a universal chemical state machine which can standardize any chemical procedure to make laboratory research faster and more reproducible.
  • Generalizations already exist but are not translated to practical tools.  The goal here is to automate the round bottom flask using a novel computer language and physical outputs to create a chemical synthesis state machine.
  • The novel code has a graphical interface making it easy to interact with.
  • He constructed a series of robotic chemistry tools which were assembled in a fume hood and hooked up to a computer.  These modules can be improved as they discover better ways of constructing each section.
  • As proof of concept the chemical computer produced Diphenhydramine, Sildenafil, and Rufinamide all on the same machine and magnitudes faster than human controlled synthesis.
  • The programming approach is part of a natural progression of any technology, and even though the backend is complex the end product helps simplify the process of using organic chemistry to produce any molecule desired.
  • Lee postulates that with heater/chiller, filtration, phase separation, evaporation, column chromatography modules approximately 60% of organic synthesis could be covered.  Adding low temperature, solid handling, and vacuum distillation modules boosts the capability to 95% of all chemistry operations.
  • The code itself is designed to work universally for any chemical synthesis sequence, and to allow for recursion and dependencies just like normal code.
  • A color coded example of how synthesis steps are associated with the graphical code and their respective physical components.
  • Lee is starting a company called Chemify to build a generalized toolbox of functions for automated chemistry.
  • He is also working on automating the discovery process – creating an array of different products and using robots to compare and select for ideal candidates.  The method involves complex and repeated trials as well as chemical auditing which would take too much brute force effort for humans to reasonably complete but are possible for automated systems such as the chemputer.
  • The original intent for this automation was to look at the origin of life.  To that end, Lee had previously constructed an evolutionary engine to assess how the earliest cells were created.  He was able to produce rudimentary cellular structures in the lab and get them to evolve in a limited fashion.
  • By automating the discovery process, more samples can be assessed and discovery experiments can progress faster than if run by a human.  Furthermore, it removes human bias from the discovery process and may yield unexpected breakthroughs.
  • The chemputer can not only perform organic covalent chemistry, but also mechanical bond chemistry.  It can assemble complex components of large molecular machines and then assemble them automatically.  Thus the chemical state machine creates a molecular machine which leads to a feedback loop.
  • All these machines use the same code and can integrate seamlessly with each other.  The ideal goal is to focus on making molecules which are universal, performing tasks which are universal, and developing methods which are universal.  Experiment replication should become effortless with such a system in place.  This system should run reliably on any compatible hardware.


What are the long term goals for this group?

  • Currently rerouting existing reactions, figuring out multiple pathways to any molecule.  In the future – inventing new reactions for highly complex molecules.


Is cloud computing in the works for chemical computers?

  • Yes, Lee is starting a company called Chemify which hopes to be the Amazon of robotic chemistry.  You can submit encrypted code to the computer, which produces the molecules in a secure and private manner.


Can you explain the cleaning process?

  • As long as you aspirate solvent through the backbone of the machine, there’s rarely a problem unless you have a precipitate or extremely sticky substance.  It is also capable of doing a strong acid wash to remove all contaminants.  Lee is also playing around with the chemputer equivalent of a RAID array.


What about low-skill users who may create highly reactive intermediates?  Are there support tools to prevent accidents? 

  • It is possible to use coding to create compatibility flags for chemputer code to prevent such accidents.  There should probably be an expert mode vs. a casual user mode.


Are you expanding the concept of universality, in terms of using solids?

  • The intent of the chemputer is to be universal for liquids and solids.  Sublimation can help with dealing with solid chemistry.


Can these reactions be “chipified”?

  • Yes, however the current iteration is designed for discovery.  Chemputers for industrial processes may be more scale optimized when designed for production rather than research purposes.  Optimal discovery size appears to be at the milliliter scale due to surface area interactions.  


What are the regulatory burdens you are facing?

  • Lee has spoken to the FDA and NIH about proving credentials.  Blockchain is an option.  The regulatory problem is likely a long way away.  Local pharmacies with chemputers could prepare drugs and chemicals for local populations via a cloud system.


How much does the setup cost?

  • Around 50-60 thousand pounds.  The main costs are the chiller and rotovape, as well as some pumps and valve heads.




Seminar summary by Aaron King.