---
license: mit
---
# gnomAD variants and GPN-MSA predictions
For more information check out our [paper](https://doi.org/10.1101/2023.10.10.561776) and [repository](https://github.com/songlab-cal/gpn).

## Querying specific variants or genes

- Install the latest [tabix](https://www.htslib.org/doc/tabix.html):  
  In your current conda environment (might be slow):
  ```bash
  conda install -c bioconda -c conda-forge htslib=1.18
  ```
  or in a new conda environment:
  ```bash
  conda create -n tabix -c bioconda -c conda-forge htslib=1.18
  conda activate tabix
  ```
- Query a specific region (e.g. BRCA1), from the remote file:  
  ```bash
  tabix https://huggingface.co/datasets/songlab/gnomad/resolve/main/scores.tsv.bgz 17:43,044,295-43,125,364
  ```
  The output has the following columns:  
  | chrom | pos | ref | alt | GPN-MSA score |  
  and would start like this:   
  ```tsv
  17      43044304        T       G       -5.10
17      43044309        A       G       -3.27
17      43044315        T       A       -6.84
17      43044320        T       C       -6.19
17      43044322        G       T       -5.29
17      43044326        T       G       -3.22
17      43044342        T       C       -4.10
17      43044346        C       T       -2.06
17      43044351        C       T       -0.33
17      43044352        G       A       2.05
  ```
- If you want to do many queries you might want to first download the files locally
  ```bash
  wget https://huggingface.co/datasets/songlab/gnomad/resolve/main/scores.tsv.bgz
  wget https://huggingface.co/datasets/songlab/gnomad/resolve/main/scores.tsv.bgz.tbi
  ```
  and then score:
  ```bash
  tabix scores.tsv.bgz 17:43,044,295-43,125,364
  ```

## Large-scale analysis
`test.parquet` contains coordinates, scores, plus allele frequency and consequences.
Download:
```
wget https://huggingface.co/datasets/songlab/gnomad/resolve/main/test.parquet
```
Load into a Pandas dataframe:
```python
df = pd.read_parquet("test.parquet")
```