Bibliographic record
Abstract
Accurate variant calling refinement is crucial for distinguishing true genetic variants from technical artifacts in high-throughput sequencing data. While heuristic filtering and manual review are common approaches for refining variants, manual review is time-consuming, and heuristic filtering often lacks optimal solutions, especially for low-coverage data. Traditional variant calling methods often struggle with accuracy, especially in regions of low read coverage, leading to false-positive or false-negative calls. Advances in artificial intelligence, particularly deep learning, offer promising solutions for automating this refinement process. Here, we present a Transformers-based framework for genetic variant refinement that leverages self-attention to model dependencies among variant features and directly processes VCF files, enabling seamless integration with standard pipelines such as BCFTools and GATK4. Trained on 2 million variants from the GIAB (v4.2.1) sample HG003, the framework achieved 89.26% accuracy and a ROC AUC of 0.88. Across the tested samples, VariantTransformer improved baseline filtering accuracy by 4%-10%, demonstrating consistent gains over the default caller filters. When integrated into conventional variant calling pipelines, VariantTransformer outperformed traditional heuristic filters and, through refinement of existing caller outputs, approached the accuracy achieved by state-of-the-art AI-based variant callers such as DeepVariant, despite not operating as a standalone caller. By positioning this work as a flexible and generalizable framework rather than a single-use model, we highlight the underexplored potential of Transformers for variant refinement in genomics. This study contributes a blueprint for adapting Transformer architectures to a wide range of genomic quality control and filtering tasks. Code is available at: https://github.com/Omar-Abd-Elwahab/VariantTransformer.
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 | 0.010 | 0.056 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.005 | 0.004 |
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".