Next-generation phenotyping using computer vision algorithms in rare genomic neurodevelopmental disorders.

Area of Science:

• Genetics
• Medical Imaging
• Computational Biology

Background:

• Interpreting genetic variants in heterogeneous disorders like intellectual disability (ID) is challenging.
• Novel ID syndromes require robust diagnostic approaches.

Purpose of the Study:

• To investigate the utility of algorithms in detecting facial gestalts for three novel ID syndromes.
• To assess if these algorithmic approaches can aid in interpreting variants of uncertain significance.

Main Methods:

• Facial features were extracted from patient photos using dysmorphism modeling and facial recognition algorithms.
• A hybrid model combined features to compare ID cohorts against a background population.
• The model was validated using images from individuals with Koolen-de Vries syndrome.

Main Results:

• Facial gestalts were significantly present for individuals with pathogenic variants in PACS1, PPM1D, and PHIP genes.
• Two individuals with de novo missense variants in PHIP showed significant similarity to the expected facial phenotype.
• Algorithmic analysis successfully identified previously unknown facial gestalts.

Conclusions:

• Facial photo analysis using algorithms can detect novel facial gestalts for ID syndromes.
• This approach facilitates the clinical and molecular diagnosis of rare and novel genetic disorders.
• The findings support the use of computational methods in diagnosing complex genetic conditions.