diff --git a/scripts/subset_matrix_table_to_genes.py b/scripts/subset_matrix_table_to_genes.py
new file mode 100755
index 00000000..fa045a4d
--- /dev/null
+++ b/scripts/subset_matrix_table_to_genes.py
@@ -0,0 +1,164 @@
+#!/usr/bin/env python3
+# ruff: noqa: PLR2004
+
+"""
+Takes a path to a MatrixTable and a name prefix
+Writes data into the test bucket for the dataset
+
+Optionally sample IDs and a locus can be provided to reduce the output
+If sample IDs are specified, output subset contains only those
+--format arg sets the output type (mt|vcf|both), default is MT
+
+new behaviour: `--genes` as a CLI argument, a list of whitespace-delimited ENSG IDs. Filters MT to rows/variants which
+    have at least one of the query genes present in the consequence annotations
+"""
+
+import logging
+import sys
+from argparse import ArgumentParser
+
+import hail as hl
+
+from cpg_utils.hail_batch import init_batch
+
+
+def subset_to_samples(
+    mt: hl.MatrixTable,
+    samples: set[str],
+) -> hl.MatrixTable:
+    """
+    checks the requested sample subset exists in this joint call
+    reduces the MatrixTable to a sample subset
+
+    Parameters
+    ----------
+    mt : the current MatrixTable object
+    samples : a set of samples requested for the subset selection
+
+    Returns
+    -------
+    The original MT reduced to only the selected samples, and depending on the
+    keep_hom_ref flag, only sites where that sample subset contain alt calls
+    """
+
+    missing_samples = samples.difference(set(mt.s.collect()))
+
+    if missing_samples:
+        raise AssertionError(
+            f'Sample(s) missing from subset: {", ".join(missing_samples)}',
+        )
+
+    return mt.filter_cols(hl.set(samples).contains(mt.s))
+
+
+def filter_to_gene_ids(
+    mt: hl.MatrixTable,
+    gene_ids: set[str],
+) -> hl.MatrixTable:
+    # turn the python set of Strings into a Hail Set Expression
+    hl_gene_set = hl.set(gene_ids)
+
+    # return rows where at least one of the query gene IDs is in the row annotation
+    return mt.filter_rows(hl.len(hl_gene_set.intersection(mt.geneIds)) > 0)
+
+
+def main(
+    mt_path: str,
+    prefix: str,
+    samples: set[str],
+    out_format: str,
+    genes: set[str] | None = None,
+):
+    """
+
+    Parameters
+    ----------
+    mt_path : path to input MatrixTable
+    prefix : prefix for file naming
+    samples : a set of samples to reduce the joint-call to
+    out_format : whether to write as a MT, VCF, or Both
+    genes : optional, set of Ensembl ENSG gene IDs
+    """
+
+    mt = hl.read_matrix_table(mt_path)
+
+    if samples:
+        mt = subset_to_samples(mt, samples=samples)
+
+    if genes:
+        mt = filter_to_gene_ids(mt, genes)
+
+    # write data to test output paths
+    if out_format in ['mt', 'both']:
+        matrixtable_path = f'{prefix}.mt'
+        mt.write(matrixtable_path, overwrite=True)
+        logging.info(f'Wrote new MT to {matrixtable_path!r}')
+
+    # if VCF, export as a VCF as well
+    if out_format in ['vcf', 'both']:
+        # remove GVCF INFO field if present - can't be exported to VCF
+        if 'gvcf_info' in mt.entry:
+            mt = mt.drop('gvcf_info')
+
+        vcf_path = f'{prefix}.vcf.bgz'
+        hl.export_vcf(mt, vcf_path, tabix=True)
+        logging.info(f'Wrote new table to {vcf_path!r}')
+
+    if out_format == 'ht':
+        table_path = f'{prefix}.ht'
+        mt.rows().write(table_path, overwrite=True)
+        logging.info(f'Wrote new HT to {table_path!r}')
+
+
+if __name__ == '__main__':
+    logging.basicConfig(
+        level=logging.INFO,
+        format='%(asctime)s %(levelname)s %(module)s:%(lineno)d - %(message)s',
+        datefmt='%Y-%M-%d %H:%M:%S',
+        stream=sys.stderr,
+    )
+
+    parser = ArgumentParser()
+    parser.add_argument('-i', help='Path to the input MatrixTable', required=True)
+    parser.add_argument(
+        '--out',
+        help='Full prefix for MT/VCF name\n'
+        '("output" will become output.vcf.bgz or output.mt)',
+        required=True,
+    )
+    parser.add_argument(
+        '-s',
+        help='One or more sample IDs, whitespace delimited',
+        nargs='+',
+        default=[],
+    )
+    parser.add_argument(
+        '--genes',
+        help='One or more Ensembl ENSG gene IDs, whitespace delimited',
+        nargs='+',
+        default=[],
+    )
+    parser.add_argument(
+        '--format',
+        help='Write output in this format. "both" writes MT and VCF, "HT" writes the rows from a MatrixTable',
+        default='mt',
+        choices=['both', 'mt', 'vcf', 'ht'],
+    )
+    args, unknown = parser.parse_known_args()
+
+    if unknown:
+        raise ValueError(f'Unknown args, could not parse: {unknown!r}')
+
+    assert not (
+        args.biallelic and args.keep_all_ref
+    ), 'choose one of --biallelic and --keep_all_ref'
+
+    init_batch()
+
+    main(
+        mt_path=args.i,
+        prefix=args.out,
+        samples=set(args.s) if args.s else None,
+        out_format=args.format,
+        genes=args.genes,
+    )