mkdir -p static/datasets
cd static/datasets
for f in CellCycle Church Derisi Eisen Expr Gasch1 Gasch2 Sequence SPO; do
wget https://oliveira-sh.github.io/datasets/${f}.arff
done
cd -
These .arff files are sourced from functional genomics research (Amanda Clare), annotated using FunCat and Gene Ontology (GO), ideal for hierarchical multi-label classification with tools like Clus or GMNB (ebi.ac.uk, dtai.cs.kuleuven.be). In the table below i summarized the main characteristics of the ten freely-available datasets used.
Datasets
CellCycle Tracks gene expression during yeast cell-cycle phases. Useful for modeling time-series or cell-state dependent features.
Church Focuses on gene expression under stress conditions with categorical functional labels.
Derisi Early yeast expression dataset (cDNA arrays), frequently used in clustering and hierarchical classification.
Eisen From Eisen et al.’s foundational work on clustering yeast gene expression profiles.
Expr A general gene-expression collection, potentially combining various experiments for broader analysis.
Gasch1 & Gasch2 Stress response datasets from Gasch et al., capturing diverse conditions like heat shock and osmotic stress.
Sequence Annotation based on gene sequence features—includes functional categories useful for integrative modeling.
SPO Likely relates to sporulation gene expression in yeast, useful for classification of developmental states.
| Dataset | Num of attributes | Num of classes | Number of classes per level |
|---|---|---|---|
| Cellcycle | 77 | 499 | 18 - 80 - 178 - 142 - 77 - 4 |
| Church | 27 | 499 | 18 - 80 - 178 - 142 - 77 - 4 |
| Derisi | 63 | 499 | 18 - 80 - 178 - 142 - 77 - 4 |
| Eisen | 79 | 461 | 18 - 76 - 165 - 131 - 67 - 4 |
| Expr | 551 | 499 | 18 - 80 - 178 - 142 - 77 - 4 |
| Gasch1 | 173 | 499 | 18 - 80 - 178 - 142 - 77 - 4 |
| Gasch2 | 52 | 499 | 18 - 80 - 178 - 142 - 77 - 4 |
| Pheno | 69 | 455 | 18 - 74 - 165 - 129 - 65 - 4 |
| Seq | 478 | 499 | 18 - 80 - 178 - 142 - 77 - 4 |
| Spo | 80 | 499 | 18 - 80 - 178 - 142 - 77 - 4 |
Further reading & resources
- Clus tool examples and settings (incl.
run_sc.pl,run_hsc.pl, eval classes) are available via the HMC dataset page (dtai.cs.kuleuven.be, pmc.ncbi.nlm.nih.gov). - For sequence-based functional prediction: see datasets like Sequence.arff, ideal for integrating sequence-derived features with expression data.
- Consider reading Pliakos et al. (2015) on representational power in gene features—addresses challenges like non-unique feature mappings (dtai.cs.kuleuven.be).