Summary

This is an exploration of genes that are related to cellular senescence. It found a list of genes that are not yet in the CellAge database but are likely to be involved in senescence. These genes are good candidates for experimental verification. A literature review showed that the top 3 candidates already have some evidence linking them to senescence.

See the full notebook for details.

Data

It uses data from the following sources:

• CellAge: Maintains a list of genes labeled either as pro-senescence or anti-senescence. These labels are experimentally verified. At the time of analysis, it included 153 genes that promote cellular senescence and 121 genes that inhibit it.
• Gene Ontology Database: Maintains a database of cellular components, biological processes, and molecular functions that genes are involved in. Genes are labeled with a list of terms.

Analyses

Functional Enrichment of GO terms

This is a traditional analysis that identifies Gene Ontology (GO) terms that are over-represented in genes involved in cellular senescence.

Gene senescence classifier

This analysis trained several machine learning models that classify whether a gene is involved in cellular senescence or not based on the GO terms for the gene. An ElasticNet classifier achieved 0.82 AUC. See the report.

Promoter / inhibitor classifier

This analysis aimed to classify whether a senescence-related gene promotes or inhibits senescence, based on the GO terms for the gene. It proved a more difficult task, partially because the training dataset is small. An ElasticNet classifier achieved 0.66 AUC. See report.