Consecutive low-frequency shifts in A/T content denote nucleosome positions across microeukaryotes

Nucleosomes are the basic repeating unit, each spanning ≈150bp, that structures DNA in the nucleus and their positions have major consequences on gene activity. Here, through analyzing DNA signatures across 1117 microeukaryote genomes, we discovered ≈150bp shifts in A/T content associated with nucleosome organization. Often consecutively arrayed across the genome, A/T peaks were enriched surrounding transcriptional start sites in specific clades. Most nucleosomes (both in vitro and in vivo) across eukaryotes aligned with A/T peaks, even in the presence of DNA modifications. Using artificial intelligence-based approaches, we describe DNA features associated with nucleosomes and construct a deep learning (DL) model for improved nucleosome occupancy prediction. Using this model, we found that ≈70% of "random" transfer DNA inserts from an in vivo fungal RB-TDNAseq library avoided DL predicted nucleosome-bound regions. This study reveals a eukaryote-wide strategy for generating cassettes of nucleosome-favorable DNAs that has a profound impact on nucleosome organization.