10.21203/rs.3.rs-10020713/v1DNAcrypt-AI: A genome-scale bioinformatics framework for coordinate-based cryptography using artificial intelligence
The human genome encodes billions of genomic positions shaped by evolutionary processes, forming an immense reservoir of natural biological entropy that remains incompletely exploited. Existing computational approaches to genome-based data encoding have focused primarily on localized sequence motifs, short tandem repeats, single-nucleotide polymorphisms, or artificially constructed DNA sequences, thereby failing to leverage the full structural and relational complexity inherent in genome-scale assemblies. Here, DNAcrypt-AI is presented as a bioinformatics framework that integrates genome-coordinate encoding, high-throughput sequence reconstruction, and machine learning-based sequence intelligence to harness the native complexity of human reference genome assemblies (hg19 and hg38) as a high-entropy substrate for information concealment. DNAcrypt-AI employs a genomic keyring process that distributes encoded information across randomly sampled chromosomal coordinates, which are subsequently expanded through the FAS2rDNA reconstruction pipeline into large-scale multi-FASTA corpora. Sequence-level relationships across these corpora are captured by Covary, an alignment-free and translation-aware machine learning framework, enabling coordinate-to-character resolution without explicit nucleotide composition constraints. Validation experiments demonstrate that DNAcrypt-AI reliably encodes and reproducibly recovers alphanumeric sequences ranging from 6 to 90 characters, producing high-entropy outputs while maintaining constant encoded data size independent of sequence length. These results establish the feasibility of genome-scale, AI-enabled coordinate cryptography and position the human genome as a novel, extensible computational resource for bioinformatics-adjacent applications. DNAcrypt-AI is available at https://github.com/mahvin92/DNAcrypt-AI.