PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 12814602

In-phase implies large likelihood for independent codon model: distinguishing coding from non-coding sequences.

Wei-Mou Zheng·Fang Wu
View original source ↗Export metadata ↓

Abstract

It is proven that under the independent codon model, the likelihood of a DNA coding sequence read according to the correct frame is asymptotically larger than that read with an incorrect frame. Based on this proposition, a single set of probabilities of the codon usage is enough for discriminating the six frames of coding sequences under the independent codon model. The direct coding sequence of Escherichia coli genome is taken as an example to examine the codon independency by using the mutual information and chi2 analysis. The contrast between the coding frame and the two offset frames is evident. A self-learning approach for generating training set is proposed to estimate probability parameters.

Explore related subjects

CodonEscherichia coliGenome, BacterialModels, GeneticReading FramesJournal ArticleResearch Support, Non-U.S. Gov't

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Wei-Mou Zheng, Fang Wu. 2003-07-21. In-phase implies large likelihood for independent codon model: distinguishing coding from non-coding sequences.. https://doi.org/10.1016/s0022-5193(03)00086-9

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2026

Transgene sequence codon optimization and composition determines replication competence of self-amplifying RNA.

Self-amplifying RNA (saRNA) is an emerging RNA therapeutic modality that can facilitate higher magnitude and more durable protein expression at substantially lower doses than nonreplicating mRNA. Unlike conventional messenger RNA (mRNA), alphavirus-derived saRNA must support a replicase-driven RNA amplification step in addition to translation, raising the possibility that transgene coding sequences impose sequence-level constraints on replication. Here, saRNA replication was found to be dependent on the codon composition of the transgene; multiple therapeutic transgenes were replication defective despite an intact Venezuelan Equine Encephalitis Virus (VEEV)-derived saRNA backbone. Replication defects were rescued by synonymous codon re-optimization of the same transgenes, indicating that nucleotide-level features of the coding sequence, rather than the encoded protein, govern replication competence. Comparative compositional analyses identified a distinct signature associated with productive replication, characterized by elevated GC (>53%) and GC3 (>63%) content, higher codon adaptation to human (>0.75), and reduced UpA (<43/kb) and UpU (<41/kb) dinucleotide density. Moreover, deliberate compositional perturbation of an otherwise replication-competent transgene shifted these features and abolished replication, supporting a causal and combinatorial role for sequence composition in defining saRNA replication outcome. These findings define an underappreciated constraint in saRNA therapeutics and motivate saRNA-specific payload design frameworks that incorporate alphavirus-associated compositional biases during transgene sequence optimization.

Codon↗
PubMed2026

CodonMoE: DNA language models for codon-dependent mRNA prediction.

MOTIVATION: Genomic language models (gLMs) face a fundamental efficiency challenge: one must either maintain separate specialized models for each biological modality (DNA and RNA) or develop large multimodal architectures. Both approaches impose significant computational burdens-modality-specific models require redundant infrastructure despite inherent biological connections, while multi-modal architectures demand increased parameter counts and extensive cross-modality pretraining. RESULTS: To address this limitation, we introduce CodonMoE (Adaptive Mixture of Codon Reformative Experts), a lightweight adapter that transforms DNA language models into effective RNA analyzers without RNA-specific pretraining. Our theoretical analysis establishes CodonMoE as a universal approximator at the codon level, capable of mapping arbitrary functions from codon sequences to codon-dependent RNA properties given sufficient expert capacity. Across four RNA prediction tasks spanning stability, expression, and regulation, DNA models augmented with CodonMoE significantly outperform their unmodified counterparts, with the HyenaDNA+CodonMoE series achieving state-of-the-art results using 80% fewer parameters than specialized RNA models. By maintaining sub-quadratic complexity while achieving superior performance, our approach provides a principled path toward unifying genomic language modeling, leveraging more abundant DNA data and reducing computational overhead while preserving modality-specific performance advantages. AVAILABILITY AND IMPLEMENTATION: Source code for the method and to reproduce the results is available at https://github.com/Kingsford-Group/CodonMoE.

Codon↗
PubMed2025

CaLMPhosKAN: prediction of general phosphorylation sites in proteins via fusion of codon aware embeddings with amino acid aware embeddings and wavelet-based Kolmogorov-Arnold network.

MOTIVATION: The mapping from codon to amino acid is surjective due to codon degeneracy, suggesting that codon space might harbor higher information content. Embeddings from the codon language model have recently demonstrated success in various protein downstream tasks. However, predictive models for residue-level tasks such as phosphorylation sites, arguably the most studied Post-Translational Modification (PTM), and PTM sites prediction in general, have predominantly relied on representations in amino acid space. RESULTS: We introduce a novel approach for predicting phosphorylation sites by utilizing codon-level information through embeddings from the codon adaptation language model (CaLM), trained on protein-coding DNA sequences. Protein sequences are first reverse-translated into reliable coding sequences by mapping UniProt sequences to their corresponding NCBI reference sequences and extracting the exact coding sequences from their GenBank format using a dynamic programming-based global pairwise alignment. The resulting coding sequences are encoded using the CaLM encoder to generate codon-aware embeddings, which are subsequently integrated with amino acid-aware embeddings obtained from a protein language model, through an early fusion strategy. Next, a window-level representation of the site of interest, retaining the full sequence context, is constructed from the fused embeddings. A ConvBiGRU network extracts feature maps that capture spatiotemporal correlations between proximal residues within the window. This is followed by a prediction head based on a Kolmogorov-Arnold network (KAN) using the derivative of gaussian wavelet transform to generate the inference for the site. The overall model, dubbed CaLMPhosKAN, performs better than the existing approaches across multiple datasets. AVAILABILITY AND IMPLEMENTATION: CaLMPhosKAN is publicly available at https://github.com/KCLabMTU/CaLMPhosKAN.

Codon↗