For safety assessment of drug candidates and chemicals, information about the compounds toxicological and pharmacokinetic properties are needed. We utilize the latest computation-based in silico programs, based upon artificial intelligence (AI), to predict these parameters directly from the chemical structure.
In order to begin safety assessment work in your drug project, an initial appraisal of potential ”target-driven toxicity” should be conducted. What is the biological role of your target and what could be the negative consequences for binding to it? Here, bioinformatic and other computationally-driven deep literature and database searches can be performed by our specialist toxicologist to understand and document these potential liabilities and rationalise them further in terms of risk mitigation planning. Next, is the question of potential ”chemically-driven toxicity” that can occur from the chemical properties in your drug candidate or chemical must be considered, which is also highly amenable to prediction using various in silico tools. This work is rapidly performed using the chemical structure as smiles and can start very early in discovery phase. Indeed, Today, the use of computational methods in the screening of potential structures with optimised performance, both from a safety and efficacy standpoint, has dramatically expedited the choice of structures for candidate development. In effect, in silico methods can act as a initial filter to facilitate time and cost effectivity in subsequent in vitro and in vivo studies. The in silico predictions also provide “warnings flags” for potential toxicities with high probability of occurring in vivo. This even assists as far as planning for clinical studies, giving a heads up on safety margin thinking and in safety mitigation planning in early clinical phases. Finally, computational methods can rapidly predict the ADME (Absorption, Distribution, Metabolism och Excretion) properties of a molecule, which is vital for assessing, together with pharmacological action and toxicity liabilities, if a structure will be suitable for development as a drug candidate.
At RISE, we help you by performing computer-based predictions (in silico) to determine toxicity liabilities, ADME characteristics and other physical-chemical properties of your molecules. We use many programs (such as ADMET Predictor® och Leadscope®) and are also actively developing our own in-house methods and models for a variety of interrelated endpoints. Here we have experts using machine learning (random forest, k-nearest neighbour, support vector machine) and cutting-edge techniques such as deep learning, molecular embedding, contrastive learning and multi-task learning. Models can also be constructed from your own data where appropriate.
We also perform QSAR predictions of mutagenicity according to ICH guideline M7 for impurities. If certain conditions are fulfilled, mitigation of further genotoxicity testing can be achieved.
It is an advantage to test the compounds in several programs since every program is trained using different chemical libraries and all use different computational methods (expert, statistical or machine learning, etc). Together our suites currently predict >100 toxicity liabilities, ADME and other physical-chemical properties. We collect all data in Excel format and provide a report covering the predictions. The report contains the interpretations you need to make decisions about your compounds. These interpretations are especially valuable to ranking properties of compounds within a chemical serie or between series, with a view to further development Our service also includes integration of the in silico predictions into forward thinking in project development, both with respect to a potential route for safety assessment and to the order of experimental requirements.
All of the above the tools and approaches utilized in pharmaceutically-related predictions can also be utilized in industrial chemical property predictions, providing the structure is within scope for the particular model. We are also very active in number of major scientific development programs where in silico-approaches are fore-fronting development of “new assessment methods” for risk assessment of industrial chemicals. These currently include EuToxRisk and Riskhunt3R within the EU and MistraSafechem in Sweden.