SMA Finder

SMA Finder is a tool for diagnosing spinal muscular atrophy (SMA) from short read exome, genome, or panel sequencing data. It takes a reference sequence (FASTA) and 1 or more alignment files (CRAM or BAM) as input, evaluates reads at the c.840 position of SMN1 and SMN2 to detect the most common molecular causes of SMA, and then reports whether the data indicates a complete loss of functional SMN1, and therefore a diagnosis of SMA.

SMA Finder has been tested on over 30,000 exome, genome, and panel sequencing samples from CMG/GREGoR rare disease cohorts as well as 200,000 exomes and genomes from the UK Biobank (UKBB). In these tests, SMA Finder's false positive rate was less than 1 in 200,000, its true positive rate was 28/28 (100%), and its positive predictive value (PPV) was 97%.

Limitations:
- does not report SMA carrier status or SMN1/SMN2 copy numbers
- does not detect the ~5% of cases caused by SMN1 loss-of-function mutations that do not involve the c.840 position
- requires at least 14 reads to overlap the c.840 position in SMN1 plus SMN2 in order to make a call
- was developed and tested on Illumina short read sequencing data generated using DNA extrated from whole blood and aligned using the BWA aligner. Performance on data from other sequencing technologies, sample types, and alignment pipelines is unknown.

Publication

For more information about SMA Finder, see:

Weisburd B, Sharma R, Pata V, et al. Detecting missed diagnoses of spinal muscular atrophy in genome, exome, and panel sequencing datasets. Preprint. medRxiv. 2024;2024.02.11.24302646. Published 2024 Feb 27. doi:10.1101/2024.02.11.24302646

Install

To install the latest version of SMA Finder, run: python3 -m pip install -U sma-finder

Example

Example command: sma_finder --verbose --hg38-reference-fasta /ref/hg38.fa sample1.cram Command output: ``` Input args: --hg38-reference-fasta: /ref/hg38.fa --output-tsv: sample1.smafinderresults.tsv

CRAMS or BAMS: sample1.cram

Output row #1: filenameprefix sample1 filetype cram genomeversion hg38 sampleid s1 smastatus has SMA confidencescore 168 c840readswithsmn1baseC 0 c840totalreads 174 Wrote 1 rows to sample1.smafinder_results.tsv
```

Usage

``` sma_finder --help

usage: smafinder.py [-h] [--hg37-reference-fasta HG37REFERENCEFASTA] [--hg38-reference-fasta HG38REFERENCEFASTA] [--t2t-reference-fasta T2TREFERENCEFASTA] [-o OUTPUT_TSV] [-v] cramorbampath [cramorbam_path ...]

positional arguments: cramorbam_path One or more CRAM or BAM file paths

optional arguments: -h, --help show this help message and exit --hg37-reference-fasta HG37REFERENCEFASTA HG37 reference genome FASTA path. This should be specified if the input bam or cram is aligned to HG37. --hg38-reference-fasta HG38REFERENCEFASTA HG38 reference genome FASTA path. This should be specified if the input bam or cram is aligned to HG38. --t2t-reference-fasta T2TREFERENCEFASTA T2T reference genome FASTA path. This should be specified if the input bam or cram is aligned to the CHM13 telomere-to-telomere benchmark. -o OUTPUTTSV, --output-tsv OUTPUTTSV Optional output tsv file path -v, --verbose Whether to print extra details during the run ```

Output

The output .tsv contains one row per input CRAM or BAM file and has the following columns:

Combining results from multiple samples

After running SMA Finder on many samples, it's often useful to combine the per-sample output tables into a single table. One way to do this is with the following shell command:

cd <directory with multiple SMA Finder output tsvs> combined_table_filename=combined_results.tsv head -n 1 $(ls *.sma_finder_results.tsv | head -n 1) > ${combined_table_filename} # get table header from the 1st table for i in *.sma_finder_results.tsv; do tail -n +2 $i >> ${combined_table_filename} # concatenate all tables done

Plotting combined results

A scatter plot summarizing read counts from many samples can be generated using the plot_SMN1_SMN2_scatter command:

python3 plot_SMN1_SMN2_scatter.py --format svg --format png ${combined_table_filename}

It generates plots like this one which is based on 16,626 exomes that include neuromuscular disease cohorts:

Poster from SVAR22

This poster on SMA Finder was presented at the SVAR22 conference: