Discovering Novel Ciliary Genes Through Machine Learning

Torriglia, Emilia - Irigoin, Florencia - Romanelli-Cedrez, Laura - Pazos Obregón, Flavio

Resumen:

The primary cilium is an evolutionarily conserved organelle present in most vertebrate cells and in specific neurons of invertebrates. It plays a central role in cellular signaling by sensing and transducing diverse stimuli—including mechanical, light, chemical, and thermal cues—and is involved in key biological processes such as development, chemotaxis, and certain forms of memory. Defects in primary cilium function lead to a group of severe disorders known as ciliopathies, many of which currently lack effective treatments. Genes encoding proteins required for the assembly and maintenance of the primary cilium are collectively referred to as ciliary genes. However, with few exceptions, ciliary proteins lack characteristic sequence motifs or structural domains, and the full complement of ciliary genes remains unknown. Here we apply a machine learning approach using publicly available single-cell RNA-seq data to identify novel ciliary genes. First, we compiled a reference set of known ciliary genes by selecting Caenorhabditis elegans genes annotated with Gene Ontology terms related to the primary cilium. This set was randomly divided, using 80% of the genes for model development and reserving the remaining 20% for independent evaluation. Single-cell RNA-seq data from C. elegans embryos were compiled and preprocessed. Genes with low variance across cell types were filtered out and expression values were normalized using MinMax scaling. We then focused on embryonic ciliated and non-ciliated neurons and performed dimensionality reduction using PCA followed by k-means clustering. Using the training set of ciliary genes, we identified clusters significantly enriched in these genes and selected genes located within enriched clusters in both neuronal populations. This strategy produced a list of eleven candidate genes, including five genes from the independent evaluation set. Notably, four of the remaining six genes lack functional annotation, making them strong candidates for previously unrecognized ciliary genes. By generating specific and testable biological hypotheses, these results provide a framework to guide ongoing experimental studies aimed at validating the involvement of these candidate genes in primary cilium assembly and maintenance.

Detalles Bibliográficos
2026
Agencia Nacional de Investigación e Innovación
cilia
predicción de función
Caenorhabditis elegans
machine learning
aprendizaje automático
Ciencias Naturales y Exactas
Ciencias de la Computación e Información
Ciencias de la Información y Bioinformática
Inglés
Institut Pasteur de Montevideo
IPMON en REDI
https://hdl.handle.net/20.500.12381/5685
Acceso abierto
Reconocimiento 4.0 Internacional. (CC BY)