Data driven model discovery and interpretation for CAR T-cell killing using sparse identification and latent...

Area of Science:

• Computational Biology
• Cancer Research
• Immunotherapy

Background:

• Quantitative mathematical models are crucial for understanding cell-based cancer therapy efficacy.
• Validating and interpreting these models requires precise mathematical formulations and experimental data analysis.

Purpose of the Study:

• To apply the sparse identification of non-linear dynamics (SINDy) algorithm to discover cell-cell interaction dynamics in experimental data.
• To infer key aspects of CAR T-cell and cancer cell population interactions using latent variable analysis and SINDy.
• To biologically interpret the discovered model terms in relation to CAR T-cell functions and population dynamics.

Main Methods:

• Application of the SINDy algorithm to in vitro experimental data of chimeric antigen receptor (CAR) T-cells and patient-derived glioblastoma cells.
• Integration of latent variable analysis with SINDy for inferring interaction dynamics.
• Biological interpretation of model terms related to CAR T-cell responses, binding, and density-dependent growth.

Main Results:

• The first application of SINDy to a real biological system for discovering cell-cell interaction dynamics.
• Inference of key interaction dynamics between CAR T-cell and cancer cell populations.
• Biological interpretability of model terms, revealing insights into CAR T-cell functional responses and population dynamics.

Conclusions:

• The data-driven SINDy approach provides unique insights into CAR T-cell dynamics compared to traditional model-first methods.
• SINDy has the potential to improve the implementation and efficacy of CAR T-cell therapy through enhanced understanding.
• This work demonstrates the utility of SINDy for advancing cell-based cancer therapies.