Understanding the Human Genome Language with Natural Language Models
Résumé
All genomes are made up of the letters A, G, C, and T signifying the different nucleotides in the DNA. About 500,000 human genomes have been sequenced, thereby creating a vast corpus consisting of the language of the genome, studying which will help uncover the different encoded messages, i.e., functions of the genome. We seek to make an automated mechanism to study this very language, taking inspiration from natural language models. The four nucleotide bases in the DNA molecule carry information that determines when and where proteins are generated in the body. Transcription factors bind to specific regions and activate or inhibit gene activity. The human genome contains information about the tools needed to build and maintain an organism,
as well as information about the individual’s risk of developing common illnesses. If we know the rules to the grammar for a bio-linguistic system, we can use that grammar to create new functional proteins or molecular machines. We can also read and write the manufacturing instructions for the ”body plan” of bacteria, plants,
animals, and the human body. These ideas are further established and discussed throughout the text and the
transformer model is developed based on these ideas. It is tested on the first chromosome of the human genome for different number of layers in the transformer model.
Origine | Fichiers produits par l'(les) auteur(s) |
---|