jheuschkel/cds-dataset

Dataset Card for Dataset Name

This is the dataset that was used to train SynCodonLM. It is made up of roughly 66 million CDS sequences (without introns), from roughly 35,000 species. The github of our model SynCodonLM can be found here:(https://github.com/Boehringer-Ingelheim/SynCodonLM) The Hugging Face of our learned model weights can be found here:(https://huggingface.co/jheuschkel/SynCodonLM).

Citation

If you use this work, please cite:

@article {Heuschkel2025.08.19.671089,
    author = {Heuschkel, James and Kingsley, Laura and Pefaur, Noah and Nixon, Andrew and Cramer, Steven},
    title = {Advancing Codon Language Modeling with Synonymous Codon Constrained Masking},
    elocation-id = {2025.08.19.671089},
    year = {2025},
    doi = {10.1101/2025.08.19.671089},
    publisher = {Cold Spring Harbor Laboratory},
    abstract = {Codon language models offer a promising framework for modeling protein-coding DNA sequences, yet current approaches often conflate codon usage with amino acid semantics, limiting their ability to capture DNA-level biology. We introduce SynCodonLM, a codon language model that enforces a biologically grounded constraint: masked codons are only predicted from synonymous options, guided by the known protein sequence. This design disentangles codon-level from protein-level semantics, enabling the model to learn nucleotide-specific patterns. The constraint is implemented by masking non-synonymous codons from the prediction space prior to softmax. Unlike existing models, which cluster codons by amino acid identity, SynCodonLM clusters by nucleotide properties, revealing structure aligned with DNA-level biology. Furthermore, SynCodonLM outperforms existing models on 6 of 7 benchmarks sensitive to DNA-level features, including mRNA and protein expression. Our approach advances domain-specific representation learning and opens avenues for sequence design in synthetic biology, as well as deeper insights into diverse bioprocesses.Competing Interest StatementThe authors have declared no competing interest.},
    URL = {https://www.biorxiv.org/content/early/2025/08/24/2025.08.19.671089},
    eprint = {https://www.biorxiv.org/content/early/2025/08/24/2025.08.19.671089.full.pdf},
    journal = {bioRxiv}
}

More Detail

We curated a comprehensive dataset of coding sequences (CDS) from the NCBI RefSeq database, encompassing nine major organismal groups: Archaea, Bacteria, Fungi, Invertebrate, Plants, Protozoa, Mammalian Vertebrates, Other Vertebrates and Viruses. To ensure broad taxonomic coverage and minimize sampling bias, we included only one representative CDS dataset per species, explicitly excluding sub-species and cell line-specific entries. This strategy helped prevent overrepresentation of well-studied organisms and ensured a more balanced view of codon usage across the tree of life.

To ensure high data quality, we restricted our selection to CDS datasets labeled as “reference”, which are curated and represent the most complete and accurate genomic assemblies available in RefSeq. These reference sequences are manually reviewed and serve as gold-standard annotations for genomic studies.

Given the disproportionate abundance of CDS entries in certain groups (e.g., Bacteria and Other Vertebrate), we applied stratified sampling to balance the dataset, with an intentional emphasis on mammalian vertebrates to support downstream modeling objectives. This emphasis reflects the relevance of mammalian systems in biotherapeutics and mRNA-based therapeutics, where codon optimization plays a critical role. A breakdown of dataset composition per organismal group can be seen in Figure 1a of our paper.

In total, the final dataset comprised 66,503,469 CDS entries from 34,769 unique species. We randomly partitioned the dataset into 90% training (59,853,123 sequences) and 10% testing (6,650,346 sequences) subsets. This large-scale dataset provides unprecedented diversity for codon modeling and is approximately six times larger than that used in any previously published codon language model. All sequences were validated to ensure they contained only canonical nucleotides (A, C, G, T), were divisible by three and contained no internal stop codons.