Autonomous process optimization involves the human intervention-free exploration of a range of predefined process parameters in order to improve responses such as reaction yield and product selectivity. Utilizing off-the-shelf components, we developed a closed-loop system capable of carrying out parallel autonomous process optimization experiments in batch with significantly reduced cycle times. Upon implementation of our system in the autonomous optimization of a palladium-catalyzed stereoselective Suzuki-Miyaura coupling, we found that the definition of a set of meaningful, broad, and unbiased process parameters was the most critical aspect of a successful optimization. In addition, we found that categorical parameters such as phosphine ligand were vital to determining the reaction outcome. To date, categorical parameter selection has relied on chemical intuition, potentially introducing an element of bias into the experimental design. In seeking a systematic method for the selection of a diverse set of phosphine ligands fully representative of the chemical space, we developed a strategy that leveraged computed molecular descriptor clustering analysis. This strategy allowed for the successful autonomous optimization of a stereoselective Suzuki-Miyaura coupling between a vinyl sulfonate and an arylboronic acid to selectively generate the E-product isomer in high yield.
Data driven high-throughput experimentation is enabling accelerated screening within pharmaceutical companies. New data science tools combined with machine learning are being implemented to efficiently tackle multivariate reaction optimization challenges.
Melodie Christensen, from Merck & Co., Inc., and UBC provide an overview of the use of automation in her Data Rich Experimentation (DRE) lab and her move towards autonomous reaction screening in conjunction with digitalization.
Melodie is an Associate Principal Scientist, Merck & Co., Inc. and a Ph.D. student at the Department of Chemistry, the University of British Columbia.
She has a proven track record in high-throughput experimentation platforms to support early and late stage pharmaceutical process development.
Dynamic Flowsheet Simulation of Solids Processes Journal
This work presents the fundamentals and exemplary applications of a generalized model for precipitation, aggregation and ripening processes including the formation of solid phases with two dimensions. The particle formation is governed by a widely applicable population balance approach. Solid formation processes are described via the numerically efficient Direct Quadrature Method of Moments (DQMOM), which can calculate the evolution of multiple solid phases simultaneously. The particle size distribution (PSD) is approximated by a summation of delta functions while the moment source term is approximated by a two-point quadrature. The moments to calculate the multivariate distributions are chosen carefully to represent the second order moments. Solid formation is based on the model of Haderlein et al. (2017) and is extended by a multidimensional aggregation model. Now, the influences of mixing, complex hydrochemistry and particle formation dynamics including nucleation, growth and aggregation on multiphase precipitation processes are modelled and simulated along independent dimensions with high efficiency.
Journal of Colloid and Interface Science
Flocculation performance using polyelectrolytes is influenced by critical design parameters including molecular weight, amount and sign of the ionic charge, and polymer architecture. It is expected that systematic variation of these characteristics will impact not only flocculation efficiency (FE) achieved but that charge density and architecture, specifically, can alter the flocculation mechanism. Therefore, it should be possible to tune these design parameters for a desired flocculation application.
Cationic-neutral and polyampholytic copolymers, exhibiting a range of molecular weights (103–106 g/mol), varying charge levels (0–100% cationic, neutral and anionic), and random or block copolymer architecture, were applied to dilute suspensions of silica microparticles (control) and Chlorella vulgaris. FE and zeta potential values were determined over a range of flocculant doses to evaluate effectiveness and mechanism achieved.
These different classes of copolymers provide specific benefits for flocculation, with many achieving >95% flocculation. Block copolymer flocculants exhibit a proposed, dominant bridging mechanism, therefore reducing flocculant dosage required for effective flocculation when compared to analogous random copolymer flocculants. Polyampholytic copolymers applied to C. vulgaris generally exhibited a bridging mechanism and increased FE compared to equivalent cationic-neutral copolymers, indicating a benefit of the anionic component on a more, complex, diversely charged suspension.
International Journal of Cosmetic Science
The main objective of this study was to optimize hair conditioner performance through variation of composition utilizing an automated cosmetic formulation platform and advanced characterization techniques as well as develop understanding of how performance (wet combing and wet lubrication) of hair conditioner is affected by its rheology (i.e. yield stress) and controlled breakdown of the formulations (dilution). The experimental results show that yield stress greatly impacts rheology, stability and performance of the lamellar gels for hair conditioning.
Automation is seen to play a major role in optimizing and customizing the formulation of each sample. The use of an automated formulation platform such as Chemspeed’s FLEX FORMAX makes it possible to vary the formulation composition of each sample simultaneously hereby saving time and cost.
Yield stress was engineered through formulation composition variation and that subsequently affected the rheology, stability and performance of the lamellar gels. Yield stress was found to have significant effect on the dilution viscosity breakdown of each formulation.
There exists a need for more automated formulation design and execution as it allows better control of formulation properties and development of optimized high-performance products.
The IBM and Chemspeed collaboration opens doors for a new era in synthetic chemistry with wide implications for the pharmaceutical and chemical industry.
Teodoro and his team push the boundaries of AI-driven lab-automation at IBM Research Europe and have successfully implemented predictive retro-synthesis in silico with the autonomous execution by a Chemspeed automated synthesis workstation.
Teodoro provides an overview of the project and how digitalization will enable us to accelerate drug development.
Teodoro is a Distinguished Research Scientist and Manager of the Future of Computing for Accelerated Discovery in the department of Cognitive Computing and Industry Solutions at the IBM Research Laboratory in Zurich.
Amino acid functionalised perylene bisimides (PBIs) form self-assembled structures in solution, the nature of which depends on the local environment. Using a high throughput photocatalysis set-up, we have studied five PBIs for the hydrogen evolution reaction (HER) under a range of conditions (pH and hole scavenger concentration) across 350 experiments to explore the relationship between supramolecular structure and photocatalytic activity. Using small angle X-ray scattering (SAXS), we show that photocatalytic activity is determined by the nature of the self-assembled aggregate that is formed with a correlation between the presence of charged flexible cylindrical aggregates and high levels of H2. Our work highlights the complexity of designing supramolecular photocatalysts, and also the power of tuning activity by the type of aggregate that is formed.
CSIRO RAMP / Stanford University
The outcomes of a 2018 collaboration between the RAMP Centre and researchers from the Appel Group at Stanford University were recently published in Science Translational Medicine.
Right now if you’re a Type I diabetic, you need to inject insulin at the appropriate time to process the glucose load from your food. Current products are slow to act, and have a residual release time in the body. That makes them inefficient, and inconvenient. The technology presented in this paper suggests a path towards new insulin formulations that are 1) fast acting, 2) have less residual time in the body, and 3) have a good “shelf life”. That means you could more reasonably inject your insulin at meal time (instead of 30-45 minutes ahead) for optimal effect. The result is better control of your blood sugar, and more convenience. CSIRO’s RAMP centre is proud to have made a contribution to this fantastic work by the Appel Group, and especially enjoyed hosting Anton A.A. Smith and Joseph L. Mann here in our Melbourne labs.
Insulin has been used to treat diabetes for almost 100 years; yet, current rapid-acting insulin formulations do not act quickly enough to provide good control at mealtime. In the RAMP centre, we implement high-throughput, controlled, radical polymerization techniques to generate a large library of acrylamide carrier/dopant copolymer (AC/DC) excipients (a substance formulated alongside the active ingredient of a medication, included for the purpose of long-term stabilization) designed to reduce insulin aggregation. Our top-performing AC/DC excipient candidate enabled the development of an ultrafast-absorbing insulin lispro (UFAL) formulation, which remains stable under stressed aging conditions for 25 hours, compared to 5 hours for commercial fast-acting insulin lispro formulations (Humalog). In a porcine animal model of insulin-deficient diabetes, UFAL exhibited peak action at 9 min, whereas commercial Humalog exhibited peak action at 25 min. These ultrafast kinetics make UFAL a promising candidate for improving glucose control and reducing burden for patients with diabetes.
See video footage from the RAMP centre at about 1:50, where you’ll see CSIRO and Stanford researchers working with the Chemspeed SWING XL platform housed in our labs.
European Polymer Journal
We describe the sequential incorporation of the quaternized monomer N-(2-(Methacryloyloxy)ethyl)-N,N-dimethylheptan-1-ammonium (QDM) (from 1 up to 5 mol %) into 2-((dimethylamino)ethyl methacrylate) polymer chains (PDMAEMA), via reversible addition-fragmentation chain transfer (RAFT) technique, to yield quasi-block copolymers (PDMAEMA-qb-P(DMAEMA-co-QDM)) with modified hydrophilic nature and thermo-induced self-assembly properties in aqueous solutions. This chemical modification promotes the formation of metastable nanostructures in aqueous medium. The morphological transitions were investigated by means of atomic force microscopy (AFM), dynamic light scattering (DLS), rheology and turbidimetry. The results indicated that the obtained nanostructures were stabilized by associative and cationic interactions conveyed by the quaternized moieties within the polymer chains. The size of these nanostructures could be modified as a function of molar mass, copolymer composition and temperature. The method described denotes an interesting alternative to modify the thermo-induced self-assembly behavior of PDMAEMA based copolymers employing low amounts of quaternized moieties.
March 09-11, 2020, Wotton-under-Edge UK – The meeting was organized by the Dial-a-Molecule, Directed Assembly, and AI3 Science Discovery Networks. Dial-a-Molecule’s vision is that in 20-40 years, scientists will be able to deliver any desired molecule within a timeframe useful to the end-user, using safe, economically viable and sustainable processes. Predicting the outcome of unknown reactions is a key challenge, and a key problem is lack of data, particularly on “failed” reactions. Synthesis must become a data-driven discipline.
Contribution using Chemspeed’s ISYNTH digitalizing, standardizing, accelerating automated synthesis solution: Encoding solvents and product outcomes to improve reaction prediction systems Dr. Ella M. Gale, University of Bristol