Table 1.

Summarizing overview of the various methodologies and data types used to construct digital twins across several healthcare fields.

Case Study	Aim of Digital Twin	Input Data	Methodology
Artificial Pancreas	Enhance blood glucose level monitoring and insulin delivery for individuals with diabetes, ensuring accurate predictions, noise reduction, and timely alerts without the need for frequent calibration.	Blood glucose data collected non-invasively via continuous monitoring. Calibration data for accurate glucose level predictions. Data related to glucose–insulin networks, external factors (e.g., physical activity), and additional hormones.	Signal processing algorithms for denoising and data enhancement. Bayesian inference for denoising. Least squares linear regression for data enhancement. Autoregressive modeling for future glucose concentration prediction.
Cardiac Digital Twin	Create detailed replicas of the heart (anatomical twinning) and simulate cardiac electrophysiology (functional twinning) for personalized testing and treatment strategies.	Three-dimensional heart scans from MRI. Clinical ECG measurements.	Universal Ventricular Coordinates for anatomical twinning. Mathematical models for cardiac electrophysiology. Fast-forward ECG modeling. Near real-time simulation of cardiac electrophysiology.
Single-cell Flux analysis	Integrate single-cell RNA sequencing data and metabolite fluxes to understand single-cell metabolic phenotypes, particularly in cancer research, aiding in phenotype discrimination.	Template metabolic networks. scRNA-seq datasets. Extracellular flux measurements.	scFBA model for metabolic analysis. Logical operators to calculate reaction activity scores. Constraints for metabolite exchanges.
Protein and DNA interactions	Construct protein–protein interaction networks for studying protein interactions and regulatory networks for protein–DNA interactions, enabling a deeper understanding of various biological processes and disease mechanisms.	Protein interaction data from techniques like IP-HTMS. Protein–DNA interaction data from ChIP-sequencing.	Bioinformatic analysis to identify and prioritize interactions. Integration with other genomic information for comprehensive analysis.
Clinical reports in oncology	Utilize natural language processing (NLP) to extract valuable information from clinical reports, particularly in cancer diagnosis, enabling better analysis and prediction of metastases presence over time.	Structured clinical reports from CT scans. Concatenated reports for multi-report analysis.	NLP for text processing. Machine learning models, including CNN and LSTM, for prediction. Multi-report analysis to improve accuracy.
Drug effectiveness	Identify subgroups of patients who may benefit from specific treatments during clinical trials, providing a more personalized and efficient approach to treatment evaluation.	Patient data and treatment outcomes. Variables describing patient characteristics.	Random forests, regression trees, and classification trees. Identification of variables affecting treatment effectiveness. Subgroup definition based on variables.
Drug repurposing for SARS-Cov-2	Identify existing drugs that can be repurposed for COVID-19 treatment by analyzing their interactions with the virus’s protein targets and predicting their efficacy, thereby accelerating drug discovery for the pandemic.	A total of 332 host protein targets mapped to the human interactome. Experimental and clinical trial outcomes.	AI-based algorithms for drug mapping. Diffusion algorithms for pathway analysis. Proximity algorithms for target prediction. Rank aggregation for drug prioritization.