Why DeepPumas?
Where Intelligence Meets Innovation: Healthcare Business Intelligence Redefined
What is DeepPumas?
Incorporate a wide range of information in your models
A model is only as good as the information upon which it was built. Incremental refinements to processing of specific information can be very helpful. But, new ways of integrating entirely different classes of information sets you up for truly groundbreaking work.
Learn DeepPumas by collaborating with us!
PumasAI is collaborating with organizations on specific projects to expedite providing winning solutions. Combine your domain expertise with our modeling and industry expertise for a bespoke DeepPumas learning experience where we simultaneously solve real-world problems. Contact us to explore options
Build bridges between departments
Facilitate interdepartmental collaboration by bridging the gap between scientific and machine learning modeling. Fully utilize the combined expertise of your machine learning researchers, mechanistic modelers and domain experts to achieve your goals.
Build drug development dashboards
Build complex models for challenging disease areas such as oncology and rare diseases to link early biomarkers with clinical outcomes. Facilitate real-world use by letting non-data scientists use model dashboards to to interpret their data.
Identify complex relationships quickly even with small datasets
Don’t let limited data availability in early development stop you from optimizing decisions. Accelerate clinical trials by using the known biology and mechanism of action along with DeepPumas to detect which patients might respond better and optimize dosing regimens in real-time.
Stay ahead of your competitors
DeepPumas is first-of-its kind and unparalleled in it’s ability to combine Scientific Machine Learning and non-linear mixed-effects (NLME) modeling. It drastically increases the scope of what information can be utilized together. Better knowledge for better business.
Biological & Physical Laws Meet Machine Learning!
Don’t choose between Scientific Modeling or Machine Learning. Combine them!
What is DeepPumas™?
(1) Rackauckas et al. 2020, “Universal Differential Equations for Scientific Machine Learning”, arXiv
(2) Keith et al. 2021, “Learning orbital dynamics of binary black hole systems from gravitational wave measurements”, Physical Review Research 3
(3) Lemos et al. 2022, “Rediscovering orbital mechanics with machine learning”, arXiv
Who should use DeepPumas?
Quantitative Systems Biologists
Quantitative Pharmacologists
Translational Modellers
Pharmacometricians
Biostatisticians
Health Outcomes Researchers
Clinical Research
Precision Medicine
Real-World Data
Manufacturing
Request a Call
Receive a Customized Plan
Successfully Deliver What Patients Need
Switch to Pumas Products
Seamlessly mix powerful modeling approaches
When we say seamless, we mean seamless. Use variables, covariates and random effects as inputs to a neural network that predicts a missing term in a dynamical system. Train everything in concert using maximum likelihood estimation while marginalizing out random effects.
Turn your knowledge of important interactions, constraints, or conserved quantities into the structural backbone of your models. Encode mechanistic knowledge as mathematical functions, transforms, or systems of differential equations.
Fully utilize the known context of your measurements. Repeated measurements from a single patient are more correlated than measurements across multiple patients so don’t treat your measurements as independent! Use nonlinear mixed effects to account for hierarchical sources of variability. Disentangle universally shared properties from individual properties, and individual properties from measurement noise.
Use machine learning to automatically learn predictive patterns from your data. Seamlessly embed machine learning in you mechanistic and hierarchical models to capture what you did not already know. Process complex covariates such as images, identify missing terms in your dynamical system, and discover individualizable transformations that relates latent variables to measurements.
