Research map / Architectures and efficiency / Attention and transformers
Extending Sequence Length is Not All You Need: Effective Integration of Multimodal Signals for Gene Expression Prediction
Zhao Yang, Yi Duan, Jiwei Zhu, Ying Ba, Chuan Cao, Bing Su
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 |
|---|---|---|---|
| add backdoor adjustment over 2 background states (n=2 vs n=0) (ablation) | Prism · K562 | MSE: 0.1863 -> 0.1789 | Improved |
| use a single background state (n=1) instead of none (ablation) | Prism · K562 | MSE: 0.1863 -> 0.1891 | Hurt |
| raise number of background states from 2 to 4 (ablation) | Prism · K562 | MSE: 0.1789 -> 0.1762 | Improved |
| raise number of background states from 2 to 6 (ablation) | Prism · K562 | MSE: 0.1789 -> 0.1857 | Hurt |
| set intervention weight alpha to 10.0 instead of 1.0 (ablation) | Prism · K562 | MSE: 0.1789 -> 0.1916 | Hurt |
| lower intervention weight alpha to 0.1 from 1.0 (ablation) | Prism · K562 | MSE: 0.1789 -> 0.1829 | Hurt |
| set diversity weight beta to 10.0 instead of 1.0 (ablation) | Prism · K562 | MSE: 0.1789 -> 0.1836 | Hurt |
| lower diversity weight beta from 1.0 to 0.1 (ablation) | Prism · K562 | MSE: 0.1789 -> 0.1789 | No clear effect |
| shorten Seq2Exp-soft test input from 200000 to 2500 | Seq2Exp-soft · K562 | MSE: 0.1856 -> 0.1996 | No clear effect |
| add H3K4me3 signal to the three initial signals (ablation) | Prism · K562 | MSE: 0.1789 -> 0.1719 | Improved |
| add DNase footprint signal to the three initial signals (ablation) | Prism · K562 | MSE: 0.1789 -> 0.1794 | No clear effect |
| load pre-trained Caduceus weights, 200k input (ablation) | Caduceus w/ signals · K562 | Pearson correlation: 0.8630 -> 0.8743 | Improved |
Closest papers by meaning
- Learning to Discover Regulatory Elements for Gene Expression Prediction (ICLR 2025 · similarity 0.87)
- Multi-modal Transfer Learning between Biological Foundation Models (NeurIPS 2024 · similarity 0.86)
- What DNA Foundation Models Learn Beyond Sequence Composition (NeurIPS 2026 · similarity 0.85)
- Context-Aware Regularization with Markovian Integration for Attention-Based Nucleotide Analysis (NeurIPS 2025 · similarity 0.85)
- GeneZip: Region-Aware Compression for Long Context DNA Modeling (NeurIPS 2026 · similarity 0.85)
- Enhancing Transcription Factor Prediction through Multi-Task Learning (Student Abstract) (AAAI 2024 · similarity 0.83)
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.
