| Popis: |
Epigenetics underpins the regulation of most genes, including many known to play a key role in the adaptive and innate immune system (AIIS). In this thesis, I developed a method called EpiNN that leverages epigenetic data to perform supervised gene classification with the goal of detecting AIIS-relevant regions of the genome. I trained EpiNN using a handcrafted list of genes with immune cell-specific epigenetic signatures and used it to scan the human genome for new AIIS-relevant regions. I detected 2,765 putative AIIS loci, of which 1,571 are uncharacterised by Gene Ontology (GO). I used stratified linkage disequilibrium (LD) score regression coupled with summary statistics from genome-wide association studies (GWAS) for 176 traits (average N=262k) to test whether the AIIS annotation is predictive of regional heritability for these traits. I detected significant heritability effects (average |τ∗| = 1.65, p < 0.05/176) for 20 out of 26 immune-relevant traits. In a meta-analysis of 63 independent heritable traits, immune-relevant traits and diseases were 4.45× (SE 0.09, p = 9 × 10−92) more enriched for heritability than other traits. The EpiNN annotation was also among the most significantly depleted in squared trans-ancestry genetic correlation in S-LDXR analysis of GWAS summary statistics from East Asian and European populations, indicating ancestry-specific effects. I analysed several of the predicted genes and found novel AIIS-specific enhancers and a novel AIIS-specific transcription start site in DNMT1, which was confirmed experimentally using PCR and Western Blot analyses. These results underscore the promise of leveraging supervised learning algorithms and large epigenetic data sets to detect genomic regions implicated in specific classes of heritable traits and diseases. |
