Digital Twin Generators for Disease Modeling

Nameyeh Alam,Jake Basilico,Daniele Bertolini,Satish Casie Chetty,Heather D'Angelo,Ryan Douglas,Charles K. Fisher,Franklin Fuller,Melissa Gomes,Rishabh Gupta,Alex Lang,Anton Loukianov,Rachel Mak-McCully,Cary Murray,Hanalei Pham,Susanna Qiao,Elena Ryapolova-Webb,Aaron Smith,Dimitri Theoharatos,Anil Tolwani,Eric W. Tramel,Anna Vidovszky,Judy Viduya,Jonathan R. Walsh

2024-05-03

Abstract:A patient's digital twin is a computational model that describes the evolution of their health over time. Digital twins have the potential to revolutionize medicine by enabling individual-level computer simulations of human health, which can be used to conduct more efficient clinical trials or to recommend personalized treatment options. Due to the overwhelming complexity of human biology, machine learning approaches that leverage large datasets of historical patients' longitudinal health records to generate patients' digital twins are more tractable than potential mechanistic models. In this manuscript, we describe a neural network architecture that can learn conditional generative models of clinical trajectories, which we call Digital Twin Generators (DTGs), that can create digital twins of individual patients. We show that the same neural network architecture can be trained to generate accurate digital twins for patients across 13 different indications simply by changing the training set and tuning hyperparameters. By introducing a general purpose architecture, we aim to unlock the ability to scale machine learning approaches to larger datasets and across more indications so that a digital twin could be created for any patient in the world.

Machine Learning

What problem does this paper attempt to address?

The paper aims to address the issues of clinical trials and personalized medical treatment recommendations in the medical field. Specifically, the paper proposes a new neural network architecture—Digital Twin Generators (DTGs)—for creating digital twin models of patients. These digital twin models can simulate the development trajectory of an individual's health condition, thereby improving medical decision-making on multiple levels:

1. **Improving Clinical Trial Efficiency**: By generating accurate digital twin models of patients, more efficient clinical trials can be designed, reducing trial costs and time while increasing statistical power.
2. **Personalized Treatment Selection**: Utilizing the patient's digital twin model, doctors can predict the effects of different treatment plans, thereby selecting the best treatment plan for the patient.
3. **Multi-Disease Applicability**: The paper demonstrates that the same neural network architecture can generate accurate digital twin models for 13 different diseases, requiring only hyperparameter adjustments. This indicates that the method has broad applicability and scalability.

The core contribution of the paper is the proposal of a generative model architecture based on the Neural Boltzmann Machine (NBM). This architecture can learn complex multivariate time series from large-scale, heterogeneous clinical datasets and generate reliable probabilistic predictions. This method not only addresses the issue of missing data but also captures the complex temporal dependencies in clinical data. Through this approach, the researchers hope to unlock the potential of machine learning technologies on larger-scale datasets, thereby advancing medical research.