Research map / Architectures and efficiency / CNN design
Reverse-Complement Equivariant Networks for DNA Sequences
Vincent Mallet, Jean-Philippe Vert
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| set Irrep ratio a/(a+b) to 0.75 instead of other ratios (0 to 1) (ablation) | Irrep · Binary TF tasks, validation | AuROC: 0.75 | Improved |
| use k-mer length k=3 instead of k in {1,2,4} (ablation) | Irrep and Regular · Binary TF tasks, validation | AuROC: 3 | Improved |
| replace Standard with RCPS equivariant architecture | CNN · Binary TF tasks | AuROC | Improved |
| replace RCPS with Best Equivariant (Irrep/Regular) | RCPS · Binary TF tasks, full data | AuROC | Improved |
| train on 1,000 sequences instead of full data; Best Equivariant vs Standard gap | Best Equivariant · Binary TF tasks | AuROC gap vs Standard: 1% vs 0.3% | Improved |
| equivariant models vs Standard for convergence speed | equivariant models · Binary TF tasks | convergence speed | No clear effect |
| ensemble of two Irrep models instead of single Best Equivariant | Best Equivariant · Binary TF tasks | AuROC | Improved |
| post-hoc averaging or ensembling two Standard models | Standard · Binary TF tasks | AuROC | Improved |
| post-hoc Standard vs Ensemble Standard | Standard · Binary TF tasks CTCF, SPI1, MAX | AuROC | Mixed |
| Ensemble Irrep vs post-hoc and Ensemble Standard at two predictions | Ensemble Irrep · Binary TF tasks | AuROC | Improved |
| Irrep (k=3, a/(a+b)=75%) vs RCPS on profile task, full data | Irrep · ChIP-nexus profiles (Oct4, Sox2, Nanog, Klf4), full data | Spearman correlation | Hurt |
| Irrep vs RCPS on profile task, 1,000 training sequences | Irrep · ChIP-nexus profiles, reduced | Spearman correlation | Improved |
Closest papers by meaning
- Caduceus: Bi-Directional Equivariant Long-Range DNA Sequence Modeling (ICML 2024 · similarity 0.85)
- Revisiting Convolution Architecture in the Realm of DNA Foundation Models (ICLR 2025 · similarity 0.83)
- HeMeNet: Heterogeneous Multichannel Equivariant Network for Protein Multi-task Learning (AAAI 2025 · similarity 0.82)
- Curriculum Learning for Biological Sequence Prediction: The Case of De Novo Peptide Sequencing (ICML 2025 · similarity 0.82)
- Levenshtein Distance Embedding with Poisson Regression for DNA Storage (AAAI 2024 · similarity 0.82)
- Equi-mRNA: Protein Translation Equivariant Encoding for mRNA Language Models (NeurIPS 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
