mRNALocator-imb: mRNA Subcellular Localization Prediction Data

mRNALocator-imb is an ensemble framework for predicting eukaryotic mRNA subcellular localization, designed to handle severe class imbalance. The model integrates physicochemical property–based features with k-mer frequency vectors and uses a hybrid Random Forest and Transformer architecture. The dataset, last updated in May 2026, accompanies the research paper and is shared under a CC-BY-4.0 license.

Use Cases

• Benchmarking new mRNA localization predictors against the described ensemble framework.
• Training models for imbalanced classification based on physicochemical property and k-mer frequency features.
• Developing hybrid architectures that combine tree-based models and neural networks for sequence analysis.

Strengths

• The accompanying research demonstrates the model outperforms conventional methods, particularly under severe class imbalance.
• The framework explicitly addresses imbalance with techniques like Random Oversampling, LDAM loss, and SMOTE.
• Data is shared under a permissive CC-BY-4.0 license, facilitating reuse.

Limitations

• The dataset is very small at 30.1 KB, suggesting limited scope or it may be a supplementary document rather than a primary data file.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for training models.

Provenance

Source

figshare

Collection Method

Likely contains sequence-derived features (physicochemical properties, k-mer frequencies) generated for the research.

Freshness

Last updated 2026-05-19 22:01:08; freshness should be verified.