Year: 2024
Type: article
Abstract: DNA/RNA are awesome and thrilling molecules visible as incredible mechanical residences which include power, extensibility, and light seriousness.But, to this point, restricted fashions are to be had to absolutely discover sequence belongings relationships for evaluation and design.Here a custom generative large language version is proposed to enable the layout of new DNA/RNA sequences to satisfy complex combinations of target mechanical residences.The model, pre-trained on a massive set of DNA/RNA sequences, is fine-tuned on 1,000 main DNA/RNA sequences for which related fiber degree mechanical houses exist, to provide an end a give up forward and inverse generative approach this is carried out in a multi-agent (achievement plan(s)/way(s) of attaining desires).Performance is tested/evaluated through (i) An (quiet, less expensive toy / new, unexpected high-quality) evaluation and DNA/RNA kind class for created DNA/RNA sequences through primary nearby matching up in an instant line search device (BLAST) search.(ii) Belongings (method of figuring out the well worth, quantity, or nice of something) and evaluation with almost identical sequences.(iii) Evaluation of resulting molecular structures.(iv) A defined/explained series repeating concept evaluation.This work creates DNA/RNA sequences with asset combinations that don't exist in nature and develops deeper expertise of the mechanistic roles of collection styles in (accomplishing or gaining with attempt) (something that hangs over, affects, and consists of each part of something) key mechanical houses (elastic modulus (E), power, longevity, failure pressure).The model provides a (producing lots with little or no waste) method to enlarge the DNA/RNA dataset, assisting in additional sequence shape analyses of DNA/RNA, and establishes a foundation for (produced through humans/not clearly occurring) DNA/RNA layout and optimization.
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: gold
Funder: Cancer Research Institute
OpenAlex ID: W4401485416