This advice is spot on!

I haven't worked on this problem, but others in my graduate lab did. If you're interested in a tool that automates some of this process (takes VCF as input; filters variants based on frequency; you'll need to map disease symptoms/phenotypes to Human Phenotype Ontology [1] identifiers), some of my former lab mates developed a web tool [2]: https://amelie.stanford.edu/submit

[1] https://hpo.jax.org/app/

[2] https://www.medrxiv.org/content/10.1101/2020.12.29.20248974v...