README.md

---
license: other
tags:
- biology
- genomics
- variant-classification
---

# The P-KNN framework source code is licensed under the MIT License. However, these precomputed scores are derived from dbNSFP. 
# Users are strictly bound by the [dbNSFP licensing terms](https://www.dbnsfp.org/license). For commercial use, you must obtain a commercial license directly from dbNSFP.


# P-KNN Precomputed Scores Dataset
This dataset provides **precomputed pathogenicity prediction scores** generated by the P-KNN method using **dbNSFP v5.2** (academic or commercial version) with joint calibration.  
It contains pathogenicity assessments for **all missense variants**, organized into multiple subfolders.
## Dataset Structure
### 1. `precomputed_score_academic_chromosome`
Includes precomputed scores derived from the **academic version** of dbNSFP v5.2a, organized by genomic coordinates.
### 2. `precomputed_score_academic_gene`
Includes precomputed scores derived from the **academic version** of dbNSFP v5.2a, organized by MANE Ensembl transcripts. You can explore these scores using the [P-KNN-Viewer](https://huggingface.co/spaces/brandeslab/P-KNN-Viewer).
### 3. `precomputed_score_commercial_chromosome`
Includes precomputed scores derived from the **commercial version** of dbNSFP v5.2c, organized by genomic coordinates.
### 4. `precomputed_score_commercial_gene`
Includes precomputed scores derived from the **commercial version** of dbNSFP v5.2a, organized by MANE Ensembl transcripts. You can explore these scores using the [P-KNN-Viewer](https://huggingface.co/spaces/brandeslab/P-KNN-Viewer).

> **Note:** For any **commercial use** of dbNSFP data, please review the [dbNSFP commercial license requirements](https://www.dbnsfp.org/license).

Both folders share the same file structure and columns. Files are typically named by genome build, variant type, database version, and chromosome, e.g.: P_KNN_hg38_missense_dbNSFP_chrY.csv
## Column Descriptions
### files in `_chromosome` folder: 
Each CSV file contains the following **8 columns**:
| Column | Description |
|--------|-------------|
| `#Chr` | Chromosome number or identifier (e.g., 1–22, X, Y). |
| `Start` | 1-based genomic start position of the variant. |
| `End` | 1-based genomic end position of the variant. For SNVs, this is typically the same as `Start`. |
| `Ref` | Reference allele observed in the reference genome. |
| `Alt` | Alternate allele observed in the variant call. |
| `P_KNN_posterior_probability(pathogenic)` | Posterior probability that the variant is **pathogenic**. |
| `P_KNN_posterior_probability(benign)` | Posterior probability that the variant is **benign**. |
| `P_KNN_log_likelihood_ratio(evidence_strength)` | Log-likelihood ratio corresponding to the **ACMG guideline evidence strength**. |
### files in `_gene` folder:
| Column | Description |
|--------|-------------|
| `genename` | Official gene symbol where the variant is located. |
| `Ensembl_transcriptid` | Ensembl transcript identifier associated with the variant. |
| `Pos(cDNA)` | Position of the variant in the complementary DNA (cDNA) sequence of the transcript. |
| `#Chr` | Chromosome number or identifier (e.g., 1–22, X, Y). |
| `Pos(hg38)` | 1-based genomic start position of the variant in the human genome build hg38. |
| `Ref` | Reference allele observed in the reference genome. |
| `Alt` | Alternate allele observed in the variant call. |
| `Pos(protein)` | Position of the variant in the protein sequence. |
| `aaref` | Reference amino acid at the given protein position. |
| `aaalt` | Alternate amino acid resulting from the variant. |
| `P_KNN_posterior_probability(pathogenic)` | Posterior probability that the variant is **pathogenic**. |
| `P_KNN_posterior_probability(benign)` | Posterior probability that the variant is **benign**. |
| `P_KNN_log_likelihood_ratio(evidence_strength)` | Log-likelihood ratio corresponding to the **ACMG guideline evidence strength**. |

## Other Folders
### 3. `dataset4commandline`
Contains the **default calibration** and **regularization** datasets used by the P-KNN command-line tool.  
For more details, see the [P-KNN GitHub repository](https://github.com/Brandes-Lab/P-KNN).
### 4. `manuscript_dataset`
Contains additional datasets used in the study:  
> *P-KNN: Maximizing variant classification evidence through joint calibration of multiple pathogenicity prediction tools.*
### 5. `dataset4viewer`
Contains the re-organized data used in [P-KNN precomputed score viewer](https://huggingface.co/spaces/brandeslab/P-KNN-Viewer).
## Citation
If you use this dataset, please cite the relevant publication (when available) and the P-KNN tool.
## Related Resources
- **Tool & Source Code**: [Brandes-Lab/P-KNN on GitHub](https://github.com/Brandes-Lab/P-KNN)
- **Gene based precomputed score viewer**: [P-KNN-Viewer](https://huggingface.co/spaces/brandeslab/P-KNN-Viewer)
- **Manuscript**: [P-KNN: Maximizing variant classification evidence through joint calibration of multiple pathogenicity prediction tools](https://doi.org/10.1101/2025.09.24.678417)
- **dbNSFP License**: [dbNSFP Commercial Use Requirements](https://www.dbnsfp.org/license)