Daneshpajouh/ChromeCRISPR: ChromeCRISPR v1.0.0 - Initial Release
Bibliographic record
Abstract
ChromeCRISPR: hybrid CNN-RNN models for predicting CRISPR/Cas9 on-target activity from sgRNA sequence. New in this release: the main model models/main_model/ holds five seeds of CNN_GRU+GC built to the architecture the article describes: three convolutional layers of 128 filters, three GRU layers of 128 hidden units, the embedding to 128, both branches reduced to 128 and concatenated to 256, GC content appended to 257, then dense layers of 128, 64 and 32. 834,497 parameters each. Scored from the checkpoints themselves: | | this model | article | |---|---|---| | Spearman | 0.8769 | 0.8760 | | mean squared error | 0.0092 | 0.0093 | The median Spearman is 0.8769 against the article's 0.8796; these runs vary less from fold to fold, which pulls the median toward the mean, and it is reported as measured. Everything was selected on a validation split drawn from the training pool: sixty candidates cross-validated five ways, each recording test_touched: false, the winner and its epoch count fixed in a hashed pre-registration, then the test set read once per seed. The isotonic calibration that returns the rank-transformed outputs to the activity scale is fitted on the validation rows shipped alongside it. Reproducing the reported results python3 scripts/verify_published_results.py All 76 reported figures agree to four decimal places: nineteen models by Spearman mean, Spearman median, MSE mean and MSE median, recomputed from the prediction vectors in artifacts/predictions/. Also included artifacts/models/CNN_GRU_GC.pth, reproducing its own prediction vector to 1.49e-07 models/retrained/ and models/published_family/, two further model sets under the same protocol, which isolate what the architecture choice is worth data/, the dataset and encoded arrays, rebuildable under a fixed seed 107 tests, including one assertion for each of the 51 architectural statements the article makes
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 | 0.004 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.004 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.130 | 0.140 |
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".