Bibliographic record
Abstract
This work presents a novel two-stage framework for RNA secondary structure prediction, introducing deep learning models that address local substructure prediction and global structure assembly. Drawing insights from RNA folding kinetics, the stage-1 (S1) model proposes local substructures as pixel-level square bounding boxes. The subsequent stage-2 (S2) models, including the scoring network (S2SC) and the encoder-decoder transformer (S2ED), leverage recurrent neural networks (RNNs) with Dynamic Programming (DP) and an encoder-decoder network with a subset prediction formulation, respectively, to achieve efficient global structure assembly. The proposed framework exhibits high accuracy on the human-transcriptome (HT) dataset. However, challenges in performance on the bpRNA dataset are observed, potentially stemming from dataset bias and limitations of the S1 model. Nevertheless, the models exhibit strong generalization capabilities compared to other deep learning models, attributed to the robust inductive bias introduced by the two-stage formulation. Anticipating advancements in computational resources and model architecture, ongoing improvements in performance are expected. The adaptability of the framework is a key highlight, as demonstrated by its capacity to be tailored to new datasets. For example, adjusting the reinforcement learning reward facilitates training the S2ED model on datasets with different structure distributions, showcasing the versatility of the proposed approach. In conclusion, this work offers a unique and effective perspective in RNA secondary structure prediction. Despite current limitations, the framework's adaptability and generalization underscore its potential for continuous evolution, particularly with the anticipated advancements in computational capabilities and the availability of diverse datasets.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".