Ablation Performance Metrics for PepLM-GNN Model Variants

Ablation experiment results compare the performance of five PepLM-GNN model variants, including the complete model and four ablated versions. The dataset contains mean and standard deviation for ACC and AUC metrics calculated from five-fold cross-validation. It was created by Ke Yan and published on figshare in March 2026.

Use Cases

• Compare ACC and AUC performance metrics across the five ablation variants (A, B, C, D, and complete PepLM-GNN) to assess component contributions.
• Analyze the standard deviation of ACC and AUC across cross-validation folds to evaluate model stability for each variant.
• Benchmark the performance of the complete PepLM-GNN model against its ablated versions (e.g., lacking GAT, GCN, or GIN modules) using the provided metrics.

Strengths

• Performance metrics (ACC, AUC) are reported as mean ± standard deviation, providing a measure of result reliability.
• Results are derived from a five-fold cross-validation procedure, a standard method for robust performance estimation.
• The dataset is clearly structured to compare five distinct model architectures in a controlled ablation study.

Limitations

• The dataset is very small at 5.5 KB, indicating it likely contains only summary statistics, not the underlying raw data or model predictions.
• Without access to the raw benchmark set or model code, the results cannot be independently reproduced or extended.
• The specific number of data points (rows) used to calculate the metrics is unknown, limiting understanding of the experiment's scale.

Provenance

Source

figshare

Collection Method

Ablation experiment performance results from a five-fold cross-validation on a benchmark set.

Freshness

Last updated March 2026.