Advances in science, technology, and information management can help us do things faster and smarter than was previously possible. Our experimental work will benefit from using effective ways of collecting and analyzing laboratory data. We can save time and money, two precious commodities, by using predictive computer programs to generate trustworthy estimates of chemical and toxicological properties. We can formulate much better products than those put on the market years ago if we proceed logically and understand the properties and functions of every ingredient being used.
Historically there are very few rules to follow in personal care formulation. Look at undergraduate textbooks in physics, chemistry, and biology. The physics book will be loaded with equations. The chemistry book will have some equations and quantitative relationships, but relatively more descriptive information. Biology is by far the least mathematically rigorous and most descriptive of the sciences. Personal care formulation resembles biology. Outside of HLB theory or Stoke’s Law, or a formula for neutralizing fatty acids and polymers, there are few set rules to guide the formulator.
Rigor has crept into cosmetic science over the years. Where once a material was either polar or nonpolar, we now have Hanson Solubility Parameters and Clog P. We know more about formulating to deliver actives, and how the polarity of the oil phase impacts the SPF of a sunscreen. Increasing the stability of a sunscreen molecule exposed to UV radiation using triplet quenching involves nothing less than quantum mechanics.
Let’s examine three examples of smarter ways of working: analyzing multiple variables using phase diagrams, using SMILES notation as an input tool to predict chemical properties using computer programs, and following the reasoning of a decision tree to establish the safety of aroma chemicals.
Analyzing Multiple Variables Using Phase Diagrams
Trial and error is the worst approach to a formulating challenge. Organizing data graphically can provide a roadmap and possibly reveal unexpected relationships. Multiple variables especially benefit from the use of phase diagrams. Assume there are four variables. They can be placed at the corners of a pyramid, and every point inside represents a possible combination. The result is a quaternary phase diagram (Figure 1).
Properties such as viscosity, micelle structures, tackiness or clarity can be data points. In this example we will explore areas of microemulsion formation. Figure 1 has oil, water, a surfactant (S), and a co-surfactant (CoS). To generate data we take a specific ratio of surfactant and co-surfactant, effectively taking a triangular slice out of the pyramid. Figure 2 shows this slice, now a ternary diagram. Point P in this triangle represents 40% water, 20% oil, and 40% surfactant and cosurfactant. Titrating water into a mixture of surfactants and oil can reveal areas of clear solutions in ratios where microemulsions are spontaneously formed.
Figure 3, from an article by Graham Barker, shows the use of a phase diagram to identify micelle structures in a stearate stick.1 Barker was able to identify the amount of water and sodium stearate necessary to make a rigid composition. This clearly shows how a phase diagram is the best way to present certain types of data and to see deeply into the fundamental properties of a formulation.
SMILES Notation as an Input Tool to Predict Chemical Properties
A key assumption in using computer models for predicting chemical properties is the validity of Structure-Activity Relationships (SARs). This allows us to compare a molecule that we don’t know about to a similar one with extensive data, a process known as “read-across.” We can also raise a red flag about a molecule if we identify a structural group in it that we know is present in a problem molecule. An example of predictive behavior is knowledge of how to design for biodegradation, for example, by having oxygen or an ester linkage built into the structure. Any relationship between a known molecule and an unknown one can provide valuable guidance.
A tool exists which allows us to predict many critical properties of a molecule, even one that has never been made. It is EPI Suite, and it is available free from the EPA.2 Going back 20 years, computer programs have been available to calculate various properties of molecules, and independent workers have developed separate modules on areas such as ecological toxicity and dermal permeability. All these programs have been conveniently merged into EPI Suite (Estimation Programs Interface).
To input a specific chemical into the program, the chemical name, CAS number, and, especially, SMILES Notation, is required. SMILES (Simplified Molecular Input Line Entry System) is simply a convention for transforming a chemical structure into a form that can be typed on a key pad. An obvious issue is that a computer does not have a key for a benzene ring. The original SMILES specification was developed at the EPA in the 1980s. In 2006 IUPAC introduced the InChI as a standard for formula representation. SMILES is more intuitive than InChI and also has an extensive range of useful software that accepts its input.
SMILES notation for common chemicals can readily be found by searching the Internet. For establishing SMILES notation for a new molecule there are rules to follow. A simple example is to use upper case for aliphatics, lower case for aromatics, and a number for the beginning and end of a ring. Thus bromobenzene becomes c1ccccc1Br. In aromatic rings, heteroatoms (o, s, and n) are also lower case.
Often, a number of equally valid SMILES strings can be written for a molecule. Ethanol can be CCO, OCC, or C(O)C. Algorithms have been developed that select a single SMILES string for a given molecule out of all the possible alternatives. For propylene glycol the CAS number is 57-55-6 and the SMILES notation is OCC(O)C. Put those in EPI Suite, hit “Calculate,” go to “All Results,” and pages of data appear. It can be converted to MS Word format by pressing a button. You will find the Henry’s Law constant, estimates of solubility, biodegradation, bioaccumulation, and atmospheric oxidation, among other things. If you do the same thing with a new molecule it would give you a big jump on predicting its physical, toxicological, and ecological profile.
The HELP section of EPI Suite provides detailed descriptions of the separate modules and detailed instructions for creating and using SMILES notation. Each section is an education in itself. For example, the BIOWIN component describes the seven models used to predict biodegradation and includes extensive references and a bibliography. Figure 4 shows the opening screen of EPI Suite with the SMILES and CAS number inserted for propylene glycol.
Figure 5 shows the calculation for ECOSAR, the Ecological Structure Activity Relationship, which demonstrates the effect of the chemical on a variety of aquatic life forms. Figure 6 illustrates the output for fugacity. Fugacity provides information on where the chemical will partition in the environment, and it is an essential property for determining environmental impact.
Knowledge of fugacity, biodegradation, and aquatic toxicity combine to show a clear picture of the environmental effects of a chemical. If a molecule biodegrades quickly, its negative effects will have less chance to manifest itself. If a material is not water soluble, it will have less chance to kill fish even if it has high aquatic toxicity, since fish are unlikely to be exposed to a significant amount. Thus, the results of EPI Suite and some common sense go a long way towards forming an intelligent profile of a chemical.
Decision Tree Reasoning to Establish the Safety of Aroma Chemicals
The RIFM (Research Institute for Fragrance Materials) is responsible for providing the industry with data on fragrance safety. A complete dossier of testing can be done on a molecule for about $1,000,000. There are over 3,000 materials in the database, so a direct assault on data acquisition would cost over $3 billion. There must be a better way, and RIFM has been looking for refinements in its methods for many years.
Testing on animals is out of the question, so alternative measures must be taken. We need an approach that is quicker, less expensive, and scientifically beyond reproach. RIFM has just brought this up to date in its new criteria document, with the broad outline launched in December 2014 through a webinar.3 The criteria paper is still in press at this time.4 The basic ideas are easy enough for a layman to understand; it actually embraces a common sense approach to the subject. Some key concepts are end points, decision tree, read across, in silico, and TTC (Threshold of Toxicological Concern).
Fragrances are complex mixtures of chemicals. Ingredients can be single aroma chemicals or natural oils composed of a large number of chemicals, often in small quantities and varying from batch to batch. Only ingredients, not the total compositions, are tested for safety. When a new fragrance material enters the system, the endpoint is the place where a decision is reached on safety. Endpoints that are established by RIFM for a fragrance material are genotoxicity, repeated dose toxicity, developmental and reproductive toxicity, skin sensitization, photoirritation and photoallergenicity, local respiratory toxicity, and environmental assessment.
A decision tree is a series of questions, which serves as a roadmap through the evaluation process. Figure 7 has a partial, simplified decision tree for the first stage of a fragrance material assessment. At every juncture a question is posed, and either a decision is possible, ending the assessment process, or another step is taken.
It is first necessary to evaluate the existing data. If there is enough, we are done. If not, we proceed to “read across.” Is our new material similar to another molecule we know about? If it is similar, and there is adequate data on the other molecule, we are finished. Confidence in using read across depends on the knowledge, experience, and judgment of the toxicologist making the decision.
The next step involves in vitro screens or in silico methods using predictive toxicology programs. In silico is a general term for using computer programs rather than performing experimental work. For toxicology studies, RIFM has access to Derek, MultiCASE, Topcat, and the OECD Toolbox. RIFM has an ongoing commitment to adding new toxicology programs if they prove applicable to the types of questions posed by fragrance chemicals. If this proves sufficient, we are at the endpoint; otherwise we go to the Threshold of Toxicological Concern (TTC). TTC has been developed to handle the thousands of chemicals we can now identify in minute quantities virtually anywhere. It originated in work by the FDA in 1995 for food products and additives. It calculates a safe level for most chemicals, even for most carcinogens, present in all food and ingested over a lifetime. Certain categories like heavy metals and endocrine disruptors are excluded, as no level is deemed safe.
Natural products are an example of compounds we cannot characterize completely, so the TTC is a useful approach to handling all the pesky little chemicals present. No chemical reaction runs to 100% completion, so every synthesized chemical we use has tiny byproducts. A way was desperately needed to deal with all these materials, and TTC is the answer. TTC is fundamentally an approach for prioritizing assessment of chemicals with low-level exposures. Three broad categories of materials are identified, shown in Table 1. The allowed exposure levels for each group are based on the decision tree work of Cramer and 5 are: Cramer Group I – 1800 μg/day, Cramer Group II – 540 μg/day, and Cramer Group III – 90 μg/day.5
Table 1: Cramer classification scheme.
- Class I: Substances with simple chemical structures and for which efficient modes of metabolism exist, suggesting a low order of oral toxicity.
- Class II: Substances that possess structures that are less innocuous than class I substances, but do not contain structural features suggestive of toxicity like those substances in class III.
- Class III: Substances with chemical structures that permit no strong initial presumption of safety or may even suggest significant toxicity or have reactive functional groups.
Graphical tools (e.g., phase diagrams), computer programs (e.g., EPI Suite), and decision trees, such as that used by RIFM, make our work more efficient, cut the time and costs of development or assessment, and allow us to better predict the safety and performance of the products we use. In a world that demands safety to humans and the environment, it cannot be ignored. And, to those developing the new chemicals that are the engines of new technologies, the insight we can gain early in the process is invaluable.
- G. Barker, “Sodium stearate-based sticks: Proposed structure, Cosmet. Toil., 102(10), 71-80 (1987).
- RIFM Webinar, Assessing the Safety of Fragrance Materials: What are the Criteria?, Dec. 17, 2014.
- A.M. Api et al., “Criteria for the Research Institute for Fragrance Materials, Inc. (RIFM) safety evaluation process for fragrance ingredients,” Food Chem. Toxicol., S0278-6915(14)00481-5; doi: 10.1016/j.fct. 2014.11.014. [Epub ahead of print] (2014).
- G. Cramer et al. “Estimation of toxic hazard—a decision tree approach,” Food Cosmet. Toxicol., 16, 255-76 (1978).
About the author:
Steve Herman is President of Diffusion LLC, a consulting company specializing in regulatory issues, intellectual property, and technology development. He has been an Adjunct Professor in the FDU Cosmetic Science Program since 1993, teaching the Cosmetic Formulation Lab and Perfumery. Steve was a columnist for GCI Magazine for 17 years and wrote a book, Fragrance Applications: A Survival Guide. His SCC activities include service as Chairman of the NY Chapter in 1992 and 2013, election to Fellow status in 2002, and instructor in the Continuing Education Program.