Model-Informed Drug Discovery and Development in Preclinical Research
There are many decisions throughout R&D, often beginning with which therapeutics should enter the portfolio. The pipeline can contain first-in-class (FIC) therapeutics which have unique entities such as a new mechanism of action or are treating a new disease and/or best-in-class (BIC) therapeutics which are aiming to compete with a drug that’s already on the market. As discussed in this article, Model-Informed Drug Discovery and Development (MID3) approaches can quickly provide insight into early portfolio decisions. Here we focus on how modeling and simulation approaches can help predict BIC properties to enable lead generation and acceleration into the clinic.
Predicting Best-in-Class Properties
A mechanistic model, most often is a system of ordinary differential equations (ODEs), describing the system dynamics between a therapeutic and the relevant biological system (single disease, multiple disease, and healthy, if relevant). Mechanistic models act as a central repository of hypotheses and data that enable modelers to de-risk decisions with incomplete data and knowledge. Included in these differential equations are biophysical reactions and parameters related to the therapeutic and the disease system such as binding affinity, drug half-life, receptors per cell, autocrine and paracrine signaling, cell dynamics, etc. Modelers will perform simulations and various analyses to determine what parameters are required to be competitive and best-in-class. In addition, computational resources are scalable which enables a modeler to simulate best, worst, healthy, or disease type scenarios to gain a better understanding of the therapeutic candidate and reduce uncertainty in human dose predictions. This helps a team determine what experiments are necessary to understand and explore any uncertainty and variability. This accelerates lead generation and candidate selection and best prepares for GLP toxicology studies. Mechanistic models are widely used across various therapeutic modalities and indications, and are helpful as the complexity of the therapeutic increases. For example, one of the difficulties of designing a bispecific antibody is that you’re identifying a drug that binds to two or more target antigens, which may or may not be expressed at similar levels and may or may not be turned over at similar kinetic levels. This leads to questions such as:
- What is the optimal affinity and avidity you need to put into your bispecific antibody design in order to be able to cover both targets, attain efficacy, and have an acceptable therapeutic index?
- What format considerations can be taken?
- Do we need a long half-life? How is this balanced with target-mediated drug disposition (TMDD)?
- Do we need to worry about TMDD if there are multiple routes of TMDD?
For the reasons stated above, a mechanistic model helps teams better understand the parameters and determine which will most likely impact the required output, and what parameter ranges or values are necessary to suffice your target product profile (e.g. dose route of administration, dosing frequency).
Competitor Differentiation and Analysis Enabling Clinical Candidate Selection and De-risking In-licensing Opportunities
Modeling and simulation approaches enable competitor differentiation and analysis by benchmarking your data against a competitor or existing preclinical, in vitro, in vivo, and clinical pharmacokinetics (PK) data with semi-mechanistic links to relevant pharmacodynamics (PD) and surrogate efficacy and safety endpoints. Generally project teams agree to relevant data (in vitro, in vivo, clinical, etc.) and parameter values from in-house data and literature that will be used for benchmarking. The model can simulate your therapy against the existing competition or comparators to show if you’re competitive and under what conditions. This builds confidence around BIC questions that impact in-licensing decisions or advancing your Lead into the Clinic. You can also consider other therapies that will be developed in the future and if your drug can remain competitive.
Knowledge Gap Analysis
Reactions and parameters that are unknown and/or unable to be estimated are identified as knowledge gaps. The model is set to match known experimental conditions for in vitro functional studies, preclinical animal studies, and clinical data where available, and other parameters may be estimated through a formal global model calibration to data. The model will provide insight into what parameters matter and which ones not so much. This also allows the scientist to identify where the model may need additional attention. This allows you to determine what the next experiment is going to be and update the model with new data to further constrain the unknown parameters or reactions. Each time you recapitulate and inform the model with updated data, you can achieve more certainty on your dose predictions, which will impact your lead and clinical candidate selection and ultimately your first-in-human studies.
Technical Due Diligence for In-licensing Opportunities
Thorough, timely, and targeted analyses are game-changing advantages when assessing in-licensing opportunities. In silico technical due diligence helps assess in-licensing opportunity risk of clinical success and optimizes the Due Diligence process. Performing technical due diligence prior to Due Diligence meetings can provide the following benefits:
- Identify potential hurdles with simulated clinical trials based on disease biology and asset mechanism of action
- Identify the most critical data to narrow down discussion and support or refute claims
- Deliver knowledge gaps and inconsistencies, sensitive parameters, in silico differentiation analysis, and best-in-class predictions
- Enable rapid follow-up analysis
- Significantly optimize the due diligence process with quantitative guidance to eliminate the inefficient skimming of lab notebook entries
Example of MID3 to Predict Best-in-Class Properties
Challenge: A project team was developing a fast follower drug for an exciting target, but they were behind their competitor. The team’s candidate was a weaker binder, which most believe is indicative of a less optimal drug. They didn’t understand what their internal assays were telling them about the molecule and the disease biology. It was unclear which properties to optimize to produce a differentiated best-in-class molecule with an improved therapeutic window, and they were on the verge of killing the project.
How MID3 Helped: The team worked with Applied BioMath to develop a mechanistic PK/PD model. Using only preclinical and published data, Applied BioMath’s analysis found that although the drug appeared inferior to the competitor in in vitro assays, the model suggested that it was actually superior in human dose predictions and was best-in-class. In this case, a binder that’s neither too weak nor too tight allows you to dose less frequently and still maintain high coverage. Counter to common belief, a tighter binder is not always better. Less frequent dosing is also important for convenience. The team was able to move into clinical development faster, and with a better drug. The competitor, who was two years ahead, halted clinical trials, likely because their drug will not be competitive due to having to dose too frequently. The competitor wasted significant time and resources while the project team jumped ahead. Presently, the project team’s drug is in Phase 2, and is poised to be first-in-class and best-in-class.