Edited by: Gil Bernard Garnier, Bioprocessing Research Institute of Australia, Australia
Reviewed by: Jean-Michel Lavoie, Université de Sherbrooke, Canada; Xiaolei Fan, The University of Manchester, UK
*Correspondence: Alexei Lapkin, Department of Chemical Engineering and Biotechnology, University of Cambridge, New Museums Site, Pembroke Street, Cambridge, CB2 3RA, UK e-mail:
This article was submitted to Chemical Engineering, a section of the journal Frontiers in Chemistry.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
In this work new methods of processing bio-feedstocks in the formulated consumer products industry are discussed. Our current approach to formulated products design is based on heuristic knowledge of formulators that allows selecting individual compounds from a library of available materials with known properties. We speculate that most of the compounds (or functions) that make up the product to be designed can potentially be obtained from a few bio-sources. In this case, it may be possible to design a sequence of transformations required to convert feedstocks into products with desired properties, analogous to a metabolic pathway of a complex organism. We conceptualize some novel approaches to processing bio-feedstocks with the aim of bypassing the step of a fixed library of ingredients. Two approaches are brought forward: one making use of knowledge-based expert systems and the other making use of applications of metabolic engineering and dynamic combinatorial chemistry.
Global uncertainty over prices of petrochemical feedstocks and the desire to significantly reduce the levels of anthropogenic generation of CO2 are the two main drivers behind current rapid development of a replacement supply chain for platform molecules of the chemistry using industries (Perlack et al.,
At present the use of bio-feedstocks in product design is relatively limited, due to the small number of molecules available on the market, primarily natural oils, flavor and fragrance substances, nutraceuticals and bio-pharmaceuticals. Very few bio-derived solvents, surfactants or monomers are available at present. However, this range is expected to be rapidly expanded, offering new opportunities for product design. The emerging question is whether our existing methods of product design in formulations and other chemistry-using industries are appropriate for the new developing supply chain based on sustainable renewable feedstocks.
Our current approach to formulations design is based on heuristic knowledge of formulators that allow to select individual compounds from a library of available materials with known properties, i.e., rheology modifiers, structure-forming agents, color and fragrance substances, bio-actives etc. (Marshall and Alaimo,
We speculate that most of the compounds (or functions) that make-up the final product can potentially be obtained from a single, or very few, bio-sources. In this case, there could be an alternative path from feedstocks to products, analogous to a metabolic pathway of a complex organism. This is represented by the top path in Figure
After Corey formalized the concept of retrosynthesis in his seminal work in 1967 (Corey,
Alternatively, other approaches to product synthesis are being developed, that do not rely on the information stored in product and reaction databases. These approaches build on advances made in the fields of metabolic engineering (Rozenman et al.,
The main element of expert systems based approaches is existing chemical knowledge of a large number of compounds and reactions. Recorded in the form of on-line databases, this knowledge is in a format that allows interrogation and rule generation to be performed using expert systems. We speculate that existing chemical knowledge will now include the necessary information for expert systems to, using retrosynthetic analysis, generate (or go some way toward) synthetic routes connecting bio-feedstocks (as starting material) with a number of existing products. Also, as more bio-derived molecules with a variety of different functionalities, are added to the existing supply chain of ingredients, and the corresponding information (the molecules and the starting materials that were used in creating them) is transferred into the existing chemical knowledge, it becomes increasingly possible that synthetic routes connecting bio-feedstocks with new products having desired properties, will be found using expert systems.
Also, new approaches to processing bio-feedstocks in designing products with specific properties through advances made in metabolic engineering and dynamic combinatorial chemistry are envisioned. Within metabolic engineering metabolic pathways are assembled and optimized (by tuning the activity of the intermediate reaction steps) for the production of molecules with desired properties (Yadav and Stephanopoulos,
The aim of this paper is to conceptualize approaches to consumer product design that are not reliant on the formulator's experiential knowledge of combining known ingredients into recipes, but, rather,
In Section Methods the proposed ideas are presented. The merits and drawbacks of the proposed ideas are discussed in Section Merits and drawbacks of the proposed approaches and conclusions are drawn in Section Conclusions.
Two variants of one methodology making use of existing chemical knowledge and expert systems are suggested, one aiming for products with known composition and a second targeting known functional properties but with an unknown composition.
As shown in Figure
The idea behind retrosynthesis is to identify an optimal synthesis route connecting a desired product with a commercially available starting material by simplifying the target product through a number of disconnections (the hypothetical reverse of a synthetic step). Each precursor is then in turn examined in the same way until a suitable starting material is identified (Corey,
One approach to retrosynthetic analysis is the use of generalized reaction rules. These are procedures for evaluating reaction types and are obtained by learning from individual reactions to obtain a generalized scheme for a type of reaction (Gasteiger et al.,
A number of systems have recently made use of advanced heuristics and databases to improve the route prediction. In the past retrosynthetic analysis was limited by the requirement for an expert chemist to manually program reaction rules. Law et al. (
To date, to the best of our knowledge, expert systems based on retrosynthetic analysis are yet to be employed in development of consumer products. However, the capabilities of such expert systems have been validated in specific cases. For instance, in Law et al. (
An alternative approach to the same end would be the use of the Network of Organic Chemistry (NOC). In 1990 (Lawson and Kallies,
Though this methodology is interesting in its own right it would be desirable to extend it to the prediction of unknown reactions not contained in the network. To this end a number of approaches could be used. In statistical terminology the problem of discovering reaction not presently contained in the network equates to an edge prediction problem. One purely statistical, though potentially very powerful, approach would be the use of hierarchical networks, for a good discussion of which the reader is referred to Barabási and Oltvai (
A question equally as important as the discovery of new reactions is that of making reactions already existing within the NOC more efficient by using better catalysts. For this purpose the previously described extended reaction core approach could be used. It would however also be desirable to conduct computational screening of catalysts. To this end reference is made to Nørskov et al. (
As shown in Figure
At present, within the context of the above-mentioned approach, expert systems are able to assist with generation and evaluation (in terms of feasibility only) of the candidate products. In Chen and Baldi (
An alternative to expert systems approach involves the use of ideas developed in the fields of metabolic engineering and dynamic combinatorial chemistry. Given only the desired properties of the product are known, a pseudo evolutionary engine is used, as illustrated in Figure
The majority of components of the above-mentioned approach are in existence. Within metabolic engineering, a number of strains are often designed (PROCESS, INTERMEDIATE OUTPUT) and evaluated in terms of percentage yield of the desired product (EVALUATION). Following the evaluation, adjustments are made and often involve deletion of unnecessary or optimization of necessary genes or a combination thereof. Within dynamic combinatorial chemistry, dynamic combinatorial libraries (DCLs) are constructed (PROCESS) and evaluated in terms of functional modularity in response to an external stimuli (INTERMEDIATE OUTPUT, EVALUATION). Following the evaluation, adjustments to the construction of DCLs are made and often involve: replacement of some of the constituents within DCLs, alteration of the reversible chemistry used, a change in the external stimuli applied. For instance, in Hanai et al. (
In Nasr et al. (
Method A of the expert systems based approach is, perhaps, the easiest to implement. As the target product is known from the start, this methodology does not involve property estimation, which can be difficult to do analytically and is costly. Retrosynthetic analysis, employed as part of this methodology, can also result in the discovery of new reactions. Its main drawback is the fact that a transformation sequence, connecting given starting materials with a target product, might simply not exist, as some or all of the required chemistry may not have been carried out yet.
Method B of the expert systems based approach, however, although also dependent on a large variety of chemistry to have been done, is not constrained by the necessity of finding a transformation sequence to a given target product, but rather by a set of properties that the target product should exhibit, which somewhat liberates the search. In fact, the final output of this approach may, intriguingly, be a product that has not been considered before. The main foreseen difficulty with this methodology is the costs associated with the necessity for property estimation, either analytically or through an experiment, for each candidate product.
The main attraction of the evolution-based approach is the potential for discovery of novel products with desired properties and the potential to discover new knowledge. The biomimetic approach of evolution-based process development requires implementation of generic principles of evolutionary development, which will necessarily sample a very large space of potential process variants. This methodology depends on the ability to sample the outcomes of each evolutionary step and to make adequate decisions, both, about the new, yet unknown phenomena that took place and which could potentially be exploited, as well as about the following steps in the process evolution. As in natural evolution, the approach is not blind, but follows some generic rules. The envisioned evolutionary approach would, at the very basic level, involve an “adjustment” (mutation) step, applied iteratively, to evolve a product generating process. However, it is not unreasonable to think of the possibility of evolving a population of product generating processes, in which case selection and crossover steps would come into play. To allow the approach to converge on the optimal (near optimal) process (or population of processes) within the allocated amount of resources and/or time, adequate selection, crossover and mutation operators would need to be designed.
The evolution-based approach has the potential to not only discover novel products with desired properties, but, intriguingly, products with additional, perhaps unexpected, functions/properties. These (additional functions/properties), of course, can be undesirable and the candidate product discarded, in the context of the product sought. However, the new knowledge, thus acquired, may benefit the design of new products and, hence, should be retained. In addition to the potential to facilitate development of other products with different functionality, the data collected using the evolution-based approach could be utilized to build physical/empirical models of the underlying physical processes involved in product generation (or help improve the existing methodology).
In this short investigation new methods of processing bio-feedstocks in consumer product design were discussed. An attempt was made to conceptualize some novel approaches to processing bio-feedstocks with the aim of bypassing the step of a fixed library of ingredients. Two approaches were brought forward and discussed: one making use of expert systems and the other, evolution-based approach, making use of advances made in the fields of metabolic engineering and dynamic combinatorial chemistry. The two main components of both approaches are: generation of a number of candidate transformation sequences/process variants and properties estimation of the candidate products (second variant of the expert systems based approach and the evolution based approach). Both [components] present challenges.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Philipp-Maximilian Jacob is grateful to Peterhouse, Cambridge, as well as the Cambridge Home and European Scholarship Scheme for PhD scholarships.