Developing a cluster-based approach for deciphering complexity in individuals with neurodevelopmental differences.

Area of Science:

• Genetics
• Neuroscience
• Computational Biology

Background:

• Global developmental delay (GDD) exhibits significant genotypic and phenotypic heterogeneity, complicating the development of targeted interventions.
• The rarity of individual genetic causes for GDD hinders traditional clinical trial designs.
• Basket trials, successful in oncology, offer a potential model for GDD research but require effective patient stratification.

Purpose of the Study:

• To assess the utility of agglomerative and divisive clustering approaches for stratifying individuals with GDD.
• To identify potential patient subgroups amenable to novel therapeutic strategies, such as basket trials.
• To explore the relationship between phenotypic clusters and underlying genetic networks in GDD.

Main Methods:

• Utilized the Deciphering Developmental Disorders (DDD) cohort, comprising 6,588 individuals with GDD.
• Applied hierarchical agglomerative clustering (HAC) and k-means clustering (divisive approach) to genotypic and phenotypic data.
• Analyzed gene network and molecular function data within identified clusters.

Main Results:

• HAC identified 16 phenotypic clusters in GDD individuals, characterized by dominant phenotypes like delayed/absent speech and seizures.
• Each phenotypic cluster contained 3-12 molecularly distinct gene sub-networks with varied functions.
• K-means clustering also segregated individuals by phenotype, but identified different genetic pathways compared to HAC.

Conclusions:

• Both agglomerative and divisive clustering methods are viable for grouping GDD patients for future basket trials.
• Subdividing phenotypic clusters into molecular sub-networks may enhance treatment success.
• A combined approach of agglomerative and divisive clustering might be necessary for comprehensive GDD treatment development.