Tzung‐Chien Hsieh

This paper identifies that truncating variants in KAT6A cause two distinct neurodevelopmental disorders with opposing gene regulatory and metabolic effects, highlighting the need for variant-level stratification in therapy development.

Bone2Gene uses deep learning to classify rare bone diseases from hand radiographs, achieving high accuracy for distinct disorders but lower for overlapping ones.

The paper identifies PSMC5 variants causing neurodevelopmental disorders and shows their impact on neuronal function and immune signaling, suggesting therapeutic targets.

The paper reclassifies a variant of uncertain significance as likely pathogenic in Mowat-Wilson syndrome using next-generation phenotyping and multimodal data integration.

This paper introduces GestaltGAN, a generative model that creates photorealistic portraits of individuals with rare genetic disorders while preserving privacy.

This study defines the clinical spectrum and genotype-phenotype correlations of BCL11A-IDD by analyzing 75 patients with molecular and clinical data.

The paper analyzes clinical and mechanistic data from individuals with ACTB/ACTG1 variants to identify distinct disease phenotypes and molecular mechanisms of non-muscle actinopathies.

The paper uses next-generation phenotyping to identify Cornelia de Lange Syndrome in Nigerian children through phenotypic similarity and exome sequencing.

This paper compares ancestry-specific and multi-ancestry polygenic risk scores for LDL prediction in East Asian populations using UK Biobank and Taiwan Biobank data, finding that multi-ancestry models are more generalizable.

This study compares human and computer attention patterns when assessing images of genetic conditions using eye-tracking and deep learning saliency maps.

The paper reports a cohort of 35 individuals with CUL3 variants causing a syndromic neurodevelopmental disorder, showing loss-of-function variants disrupt protein homeostasis by impairing proteasomal degradation of CUL3 substrates.

The paper evaluates a structured diagnostic framework for ultra-rare diseases, showing improved genetic diagnoses and novel gene-disease associations using exome sequencing and AI tools.

The paper expands the phenotypic spectrum of NAA10 and NAA15-related neurodevelopmental syndromes by analyzing clinical data from 106 NAA10 and 66 NAA15 variant carriers, revealing variable severity and multi-system involvement.

GestaltMatcher improves diagnosis of ultra-rare genetic disorders by matching facial phenotypes to molecular diagnoses using a deep learning model trained on a large dataset of patient photographs.

This study reports on 25 new KBG syndrome patients, identifies common clinical features, and provides recommendations for diagnosis and treatment.

The paper introduces an AI-driven method that uses facial images to prioritize exome data, achieving 99% top-10 accuracy in identifying disease-causing genes.

This paper evaluates deep learning for syndrome recognition in metabolic disorders, finding improved accuracy with larger training sets but no significant effect from ethnicity or sex.

The paper evaluates clinical features, flow cytometry, and facial image analysis to classify GPIBDs and identify gene-specific patterns, finding that facial recognition software is most effective for gene prediction.