Model Card for SynCodonLM This repository contains code to utilize the model, and reproduce results of the paper Advancing Codon Language Modeling with Synonymous Codon Constrained Masking . Unlike other Codon Language Models, SynCodonLM was trained with logit level control, masking logits for non synonymous codons. This allowed the model to learn codon specific patterns disentangled from protein level semantics. Pre training dataset of 43 Million CDS is available on Hugging Face here. Installation Usage SynCodonLM uses token type ID's to add species specific codon context. Before use, find the token type ID (species token type) for your species of interest here! Or use our list of model organisms below Embedding a Coding DNA Sequence Codon Optimizing a Protein Sequence This has not yet been rigourosly evaluated, although we can confidently say it will generate 'natural looking' coding DNA sequences. Embedding a Coding DNA Sequence Using our Model Trained without Token Type ID Citation If you use this work, please cite: Model Organisms Species Token Type IDs Organism Token Type ID E. coli 30 S. cerevisiae 118 C. elegans 212 D. melanogaster 190 D. rerio 428 M. musculus 368 A. thalia…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy