pcbi.1014240.t004: Synthetic and Experimental Calcium Dynamics for State Classification

A dataset used to train and validate a large-kernel convolutional neural network for classifying eight distinct dynamical states of intracellular calcium ions under realistic noise conditions. The synthetic training data was generated using chemical Langevin equations, and validation achieved 96.8% accuracy on experimental data from pancreatic β-cells and other cell lines. The dataset is 5.5 KB in size, authored by Jaesung Choi, and last updated on 2026-04-29.

Use Cases

• Training neural networks to classify calcium ion dynamical states based on synthetic data with intrinsic fluctuations.
• Benchmarking classification algorithms against SVM and Random Forest baselines using experimental calcium data.
• Studying the persistence of deterministic organizational principles in stochastic biological systems.

Strengths

• Validation on experimental data achieved 96.8% accuracy, substantially outperforming SVM (54.0%) and Random Forest (51.6%) baselines.
• Synthetic training data was generated with realistic intrinsic fluctuations using chemical Langevin equations.
• The dataset is openly available under a CC-BY-4.0 license.

Limitations

• Row count and column-level documentation are unknown, which may limit suitability assessment.
• The dataset is very small at 5.5 KB, indicating limited scope.
• Last updated 2026-04-29; freshness should be verified.

Provenance

Source

figshare, author Jaesung Choi.

Collection Method

Synthetic data generated via chemical Langevin equations; experimental data from pancreatic β-cells, WT-HEK293, STIM-KO, and ORAI TKO cell lines.

Freshness

Last updated 2026-04-29 17:57:02.