An unsupervised explainable AI reveals oligonucleotide-usage zones matching 2,000 chromosome bands

Unsupervised and explainable AI can uncover genomic features beyond human expectations. Oligonucleotides such as penta- and hexanucleotides often function as core binding motifs for regulatory proteins, and their usage provides powerful tools in functional genomics. We applied an unsupervised, explainable AI to their odds-ratio profiles of all 1-Mb euchromatic fragments in the human genome. Clustering using penta- or hexanucleotide odds ratios (1,024 or 4,096 variables) unexpectedly revealed nearly 2,000 distinct zones. CG-containing penta- and hexanucleotides (244 and 1,185 variables) produced comparable ∼2,000 zones, indicating fundamental genomic segmental structures. Human chromosomes exhibit Giemsa-banding patterns comprising 850 bands at prometaphase and ∼2,000 at prophase. Using single-nucleotide coordinates available for the 850 bands, we identified diagnostic oligonucleotides distinguishing Giemsa-negative and Giemsa-positive regions. Computational pseudo-band reconstruction based on these oligonucleotides generated genome segmentations that more closely paralleled the AI-derived ∼2,000 zones than the 850 bands. These findings indicate that AI captures characteristic sequence features of chromosomal bands and can predict high-resolution banding from genome sequences alone, thereby bridging classical cytogenetics and modern AI-based genomics.