An in silico model employs computer software programmes to conduct virtual experiments. This differs from artificial intelligence (AI), which seeks to construct computer systems that simulate human problem-solving behaviour.
As opposed to in vitro experiments conducted alone, in silico models enable researchers to add nearly unlimited parameters, making their results more relevant and applicable to organisms as wholes.
Prediction of Pharmacokinetic Properties
Predicting pharmacokinetic properties is one of the key tasks involved with drug discovery. To reduce both expense and time spent experimenting on living subjects, methods must be developed that provide relevant information about potential candidates before their synthesisation; such methods must predict ADMET properties such as absorption, distribution, metabolism, excretion, and toxicity.
Utilisation of in silico models for drug development purposes has gained acceptance across healthcare organisations, and total funding for in silico clinical trials and drug testing reached £750 million by 2019 as evidence of growing trust for this technology.
While in vitro and in vivo models use actual biological materials or animals to investigate hypotheses, in silico modelling uses specially developed computer programmes for virtual experiments. These programmes can be run with various sets of parameters to create virtual patient populations as test subjects, making in silico modelling an invaluable asset in drug development costs and timeline reduction.
Predicting ADMET properties has been made easier through computational models that span from perfusion-limited models that reduce the body into compartments to describe how drugs, molecules, or nanoparticles pass through them to prediction models for intestinal absorption, Caco-2 permeability penetration, blood-brain barrier penetration, and protein binding.
These models use in silico models to calculate properties such as plasma and tissue partition coefficients, clearance rates, unbound fraction, volume of distribution, and clearance fraction. A variety of models exist that predict these parameters, from simple diffusion models through multi-compartment and three-compartment models.
Models that predict hepatic clearance and renal excretion are also being developed, which can provide useful data when assessing whether substances might interfere with hepatic enzymes such as CYP450 or P-gp and potentially cause toxicity. This field of research remains active, with predictive models still under refinement.
Prediction of Pharmacodynamic Properties
Validated in silico models provide invaluable insights for reducing the time and costs associated with drug development. Such models allow the prediction of pharmacodynamic properties such as absorption, distribution, metabolism, excretion, activity spectra, and transport toxicity. Their predictive power depends on how accurately their framework and assumptions reflect actual interactions among cells within the body.
Computational biology models that employ network structures to represent biological processes are increasingly being employed for in silico drug discovery and repurposing (NB-DRP). These models incorporate relationships among multiple biological compounds into an algorithmic graph, providing the chance to detect emerging properties at a network level as well as examine how complex relationships may contribute to disease phenotypes.
As one example, in silico models can be developed to predict how drugs interact with target and off-target receptors on cells, providing information to be used as prioritisation criteria in further experimental studies, repurposing existing drugs into new compounds, or formulating repurposed versions of existing ones. This approach reduces both clinical trial costs by decreasing the physical experimental work required as well as the number of compounds needed for clinical trials by cutting costs per compound production.
Furthermore, the in silico model can help identify potential drug interactions and adverse reactions based on predicted interactions or biological molecules; this information can then be used to inform risk assessments that assess how a new medication might impact an individual.
Predicting the pharmacokinetic properties of drugs using in silico models is one of the fastest-growing applications of in silico modelling, making a valuable contribution to drug development by eliminating animal testing as part of clinical trials and speeding up drug discovery processes. This application helps save both money and animal lives during clinical trial phases.
Medical device and pharmaceutical companies are increasingly turning to in silico models as a way to shorten development cycles without sacrificing quality or safety. Such models can predict and validate an array of pharmacodynamic properties, such as plasma protein binding, tissue/plasma distribution coefficients, and volumes of distribution.
Prediction of Biochemical Properties
In silico models can be an invaluable way of predicting biological properties that cannot be reliably measured through experimental means. While each model differs, in silico modelling usually involves using a computer programme to run simulations of the system to which the model is applied and alter various system parameters to predict how a biological system responds to drugs or treatments. Once complete, this information can then be used to predict its behaviour or response.
Mallet and de Pillis recently developed a simple model involving only two cell types that replicates both the compact-circular morphology of their simulated compact tumour as well as the wild papillary structures seen in experimental data sets. Their approach relies on hybrid cellular automata-partial differential equation technology.
A complex in silico model of cellular networks could also help identify inhibitors for specific cancer enzymes and enhance our understanding of their interactions and effects, thus increasing our chances of devising successful therapeutic strategies.
Utilising in silico models to select compounds for clinical trials saves both costs and time by quickly screening multiple potential candidates at once and helping identify which are likely to have the highest impact, optimising the research budget.
As evidence of its rising acceptance and confidence, in silico clinical trial funding surged to $750 million in 2019. This technology provides an alternative to actual human trials by simulating virtual patients using computer simulations of medical products or pharmaceuticals being tested on them.
In silico testing is an efficient and cost-cutting way to test chemical toxicity without animal-based tests; however, it requires considerable skill and regulatory insight in order to run simulations correctly and interpret results accurately.
Prediction of Toxicity
In silico models are mathematical and computational representations of biological phenomena and structures, typically using formulae, equations, or computer programmes as representations. Most models are created based on an initial biologically inspired set of update rules to define how model elements interact with one another. When new information, such as new drug compounds or biological targets, is introduced into an updated model, it will produce predictive alerts corresponding to them that can then be used as the basis of experimentation design and execution.
An updated rule-based model that accurately predicts hepatotoxicity can be used to investigate the impact of different concentrations of new drug compounds on their potential hepatotoxicity, commonly referred to as an “in silico trial.” Such experiments aim to help identify compounds that cause hepatotoxicity early on and thus reduce time, cost, and human resource requirements associated with subsequent experimental testing.
Predicting the toxic potential of drugs is an integral step in drug development, and various in silico methods have been created for this purpose. These approaches can generally be divided into four major groups: (Q) SAR, expert schemes, alliances, and read-across tools.
Chemical chronic toxicity, the toxic impact of chemicals after long-term or repeated sublethal exposures, is one of several critical endpoints that is difficult to assess using traditional experimental tests. To address this challenge, various in silico models combining chemical descriptors with machine learning algorithms have been created that can reliably distinguish between substances with and without chronic toxicity.
However, the effectiveness of these models varies considerably and therefore must be developed based on five validation principles: an endpoint; an algorithm; appropriate measures of goodness-of-fit, robustness, and predictivity; and mechanistic interpretation (if possible). Furthermore, good laboratory practises for quality data play an integral part in computational toxicology, with their appropriate application leading to safer and more effective drugs being developed for humans.