The Cellworks Biosimulation Platform
BioPharma is challenged with high failure rates in clinical trials resulting in unacceptably low FDA approval rates and correspondingly high R&D costs. A big reason for these low rates and increasing costs is the complexity of the disease, and the lack of a full understanding of the mechanism by which a drug works. Though clinical trials do attempt to select a homogeneous population, disease heterogeneity and drug complexity result in a high failure rate. Are there techniques we can bring to bear on the drug development domain from other scientific disciplines?
Most engineered products, be it a semiconductor chip, a fighter jet, or a drug-eluting stent are modeled and simulated, in software, during their design phase before they are sent into fabrication and manufacturing. This avoids a costly trial-and-error empirical approach that might otherwise be obligatory. So, why not apply the same efficiency to medicinal drug development? The general answer to this question has been, ‘biology is too complex’, or ‘it is impossible.’
The Cellworks Platform
The Cellworks platform was created precisely to minimize the trial-and-error, empirical approach used in drug development, via bio-simulation technology.
We started with the view that bio-simulation modeling of a human cell was a herculean task, but not impossible. We believe it is essential to build a bio-simulation model to make drug response predictions, and to enhance our understanding of disease physiology and medicine. A decade and over a thousand man-years later, we have made a dent in what many believed was ‘impossible’.
What does 'bio-simulation' mean? In a nutshell, bio-simulation is the ability to predict the phenotype response of a human cell to an external stimulus, such as a drug ligand or radiation. We utilize this technology, for instance, to guide us on which standard-of-care therapy a patient may respond to.
Cellworks's bio-simulation models provide a framework in designing therapeutics to transform an entirely empirical drug development science into a more precise and predictive discipline. They provide us a causative rationale as to why some therapies work, while other don’t. They also reveal the boundaries of knowledge & understanding of biological systems, which in turn guides us on where we, and others, should look next in wet-lab experiments.
Primary Components of the Cellworks Platform
Five primary components of the Cellworks Platform are a bio-simulator, pathway models, drug models, a therapy design methodology and a cloud-based infrastructure.
We use machine learning techniques at the level of statistically appropriate biochemical pathways to understand their behavior and represent them mathematically in the model. We combine these smaller pathways into larger macro interactions using a manual curation approach. The bio-simulator is built to handle differential equations and other mathematical structures, at the rate of millions per second per bio-simulation. Drug models include a quantitative model of drug targets, binding affinities, reaction rate constants coupled with pharmacokinetic and drug economic data. The therapy design engine is an intelligent combinatorial examination of drug combinations to understand phenotype response. The cloud-based infrastructure can run millions of prediction bio-simulations in parallel, and has a demonstrated ability to handle thousands of patients per day. These patients are divided between real-life clinical cases for our Precision Medicine products, and genome-bank driven drug discovery cases for our Stratified Medicine pipeline. The Cellworks platform was created to be a unified analytical representation of biological knowledge, collected from all known heterogeneous datasets, to be applied to pathologic disorders and their treatment. It is continuously enhanced as new biological knowledge is gained by researchers across the globe.
The Cellworks platform starts with the model of a healthy human cell whose physiology can be simulated at the level of biochemical pathways. Modulated by a patient’s genomic profile it creates a personalized model of the patient’s tumor cell. It then dynamically simulates the phenotype response of this tumor to various drug agents. For the desired phenotypes, by examining interactions across a huge network of signaling and metabolic pathways, the Cellworks platform reveals actionable novel therapies for an efficacious response.
Ex-vivo testing and purist machine learning technologies operate in a ‘black box’ mode, with little explanatory power. Cellworks is different. For each single patient, we describe with complete explanatory power and with full causality, the mechanism by which a therapy works for a given patient’s disease. All the inner workings of the patient’s disease pathway interactions are presented in a transparent manner to the clinician. Because, explanatory power is believing.