Entelos recognizes that validated accuracy of our PhysioLab®
platforms is important in order for our clients to trust our
predictions and recommendations.
Our unique Entelos methodology has been proven accurate time and again, with discerning global clients leveraging in silico modeling in their drug and product development processes.
The construction of an Entelos PhysioLab® model starts with dependable data sources. Primary sources are journal articles, expert reviews, and respected textbooks which may all serve as input, along with our own proprietary network of experts, consultants and collaborators. A comprehensive analysis of current,relevant physiological pathways and functions is conducted before we begin the qualitative design.
In building and applying a PhysioLab® model, Entelos and the client first agree upon the research question. The key scoping issues are: 1) which biological processes must be included in the model to answer the research question, and 2) what clinical outputs, biological “behaviors”and biomarkers or potential biomarkers must the model produce to adequately describe the healthy state, the diseased state, the treated state, and the potential for response.
Entelos then reviews the scientific literature and generates a qualitative illustration of the architecture showing all processes and relationships to be included in the model. Next, Entelos writes a large series of linked equations describing each of the included biological pathways and calibrates them against public literature or client proprietary data (note: at all stages confidentiality of client data is maintained) to ensure each is performing properly.
Validation and Research
Once the quantitated PhysioLab® model is built, calibrated, and validated, predictions can be generated by altering any of the values in any of the many equations – including those representing the critical mechanistic pathways important for disease progression and variability in therapeutic response. These changes must be consistent with literature or client data.
These alterations reflect hypothesized differences due to genetics, lifestyle, environment, level of treatment, or other factors. Depending on the research question, it is possible to simulate response across many patient types, or intensively investigate critical subpopulations of therapeutic or commercial interest. The observed ranges of values may be wide, as with all study populations. Entelos computing technology explores the mechanistic variability potentially present in real individuals – mastering the systematic evaluation of multiple mechanistic axes – an enormous task incalculable without in silico help.
Each combination of values can be thought of as a candidate virtual patient. When generating virtual patients in such large complex biosimulation models, the goal is to generate a broad variety of valid virtual patients, so that they will exhibit the full range of clinical and biomarker outputs actually observed in healthy, diseased, and treated states reported in literature or in client trials. A virtual patient is considered valid if it generates all the measurements (and measurement correlations) within ranges found in literature or client trials, simultaneously and at all available time points.
The implications of patient variability of response can be further explored using relevant data on populations of individuals afflicted with the studied condition. Starting with a suitably diverse set of valid virtual patients, Entelos can generate a series of virtual populations. These virtual populations represent phenotypically similar sets, matching clinical or epidemiological data, with different percent contributions of individual virtual patients (prevalence weighting). The summary statistics (e.g., mean response, standard deviation of response) for each hypothesized population are then compared to those for populations appearing in actual literature or client trial data. The conforming virtual populations are then ranked low-to-high based on drug response. The “best case” and a “worst case” for drug performance at population level are defined.
The contribution of each literature or client trial dataset to limit the predicted clinical phenotype of the populations can be systematically assessed by omitting each dataset to see if the best or worst case changes. With multiple virtual populations available we can, for example:
- Test a biomarker panel predicted from a single virtual population for its robustness across all virtual populations.
- Compare the best-worst range of possible summary statistics for a treated population versus a placebo population or a population treated with a competitor drug.
The end product of all this careful preparation, planning, and testing is a one-of-a-kind modeling process relied on by world-class collaborators and clients.
We invite you to read more on how the Entelos PhysioLab® models have predicted successes and failures accurately, defined mechanisms of action and identified biomarkers – streamlining or eliminating the need for trial and error in vivo studies – resulting in sizable cost savings, and time-to-market efficiencies.
Learn More About the Entelos Approach