Fig. 5

Inputting terms with the support of PheNominal versus manual extraction downstream with NLP tools. In this example, we derived several HPO terms derived from patient notes with the help of PheNominal. We provide realistic examples of 3 reproducible copy-paste errors like “limb” to “lim” that produce missing terms when parsed by NLP algorithms such as Doc2Hpo. The derived HPO terms are then passed to Phen2Gene, which ranks potential disease genes based on phenotype information alone. In this example, the rank drops from 3 to 127, which may result in a variant in a candidate gene (SPG11) not to appear on the final diagnostic report