Bioinformatics Analysis of Viral Codon Usage Bias and Host Adaptation
Introduction
Viral codon usage bias represents a fundamental genomic signature that reflects the evolutionary pressures acting on viral genomes within their host environments. The genetic code is degenerate, with 61 sense codons encoding 20 amino acids and three stop codons, and synonymous codons are not used with equal frequency in any organism [1, 2]. This non-random usage of synonymous codons, termed codon usage bias, arises from a complex interplay of mutational pressures, natural selection for translational efficiency, and host-specific constraints [3, 4]. For viruses, which are obligate intracellular parasites, codon usage patterns provide critical insights into host adaptation, replication dynamics, and immune evasion strategies [5, 6].
The analysis of viral codon usage bias has become an essential component of computational genomics in veterinary virology. By quantifying the degree to which a virus's codon usage matches that of its host, researchers can infer the evolutionary history of host switching events, predict the potential for zoonotic transmission, and identify genomic targets for vaccine development [7, 8]. This article provides a comprehensive examination of the bioinformatics methodologies employed to analyze viral codon usage bias, the biological mechanisms underlying these patterns, and their implications for understanding host adaptation in veterinary pathogens.
Biophysical and Molecular Basis of Codon Usage Bias
Translational Efficiency and tRNA Abundance
The primary selective force driving codon usage bias is the optimization of translational efficiency. Within a given host cell, the abundance of transfer RNA (tRNA) molecules varies substantially among different isoacceptor families [9, 10]. Codons that are recognized by abundant tRNAs are translated more rapidly and with greater accuracy than those recognized by rare tRNAs [11, 12]. This phenomenon, known as translational selection, results in the preferential use of "optimal" codons in highly expressed genes [13, 14].
For viruses, the reliance on host cellular machinery for protein synthesis creates a strong selective pressure to match the host's tRNA pool [15, 16]. Viruses that exhibit high codon adaptation to their host typically achieve higher rates of protein synthesis, more efficient genome replication, and increased viral fitness [17, 18]. Conversely, viruses with codon usage patterns that deviate substantially from their host may experience reduced translational efficiency, attenuated replication, and decreased pathogenicity [19, 20].
mRNA Secondary Structure and Stability
Beyond translational efficiency, codon usage influences mRNA secondary structure and stability. The folding energy of mRNA molecules, measured as Gibbs free energy (delta G), is affected by the specific nucleotide composition of codons [21, 22]. Regions of mRNA with stable secondary structures can impede ribosome scanning and reduce translation initiation efficiency [23, 24]. Viruses must balance the need for codon optimization against the requirement for appropriate mRNA structural dynamics [25, 26].
The structural constraints on mRNA folding can be visualized through computational RNA folding algorithms that generate color-coded 3D representations of predicted secondary structures [27, 28]. In these representations, regions of high thermodynamic stability are typically colored in cooler hues (blue to green), while regions of low stability appear in warmer hues (yellow to red) [29, 30]. This color coding allows researchers to identify structural elements that may regulate translation initiation, ribosomal frameshifting, or RNA stability [31, 32].
Dinucleotide Composition and CpG Depletion
Dinucleotide frequencies, particularly the frequency of CpG dinucleotides, represent another critical dimension of viral codon usage bias [33, 34]. In vertebrate genomes, CpG dinucleotides are substantially underrepresented due to the methylation-mediated deamination of cytosine to thymine [35]. Many RNA and DNA viruses exhibit similar CpG depletion patterns, which may reflect host-driven selective pressures [1, 2].
The mechanism underlying CpG depletion in viral genomes involves the host innate immune system. Pattern recognition receptors, including Toll-like receptor 9 (TLR9) and members of the zinc finger antiviral protein (ZAP) family, can detect CpG-rich RNA and DNA sequences [3, 4]. Viruses that maintain low CpG frequencies may evade these host immune surveillance mechanisms more effectively [5, 6]. This phenomenon has been demonstrated for multiple veterinary pathogens, including African swine fever virus and various coronaviruses [7, 8].
Computational Methods for Codon Usage Analysis
Codon Adaptation Index (CAI)
The codon adaptation index (CAI) is one of the most widely used metrics for quantifying the degree of codon usage bias in viral genomes [9, 10]. CAI measures the similarity between the codon usage of a target gene and a reference set of highly expressed genes from the host organism [11, 12]. The index ranges from 0 to 1, with values approaching 1 indicating strong adaptation to the host's codon usage preferences [13, 14].
The calculation of CAI involves several steps. First, a reference set of highly expressed host genes is identified, typically based on transcriptomic data or ribosomal profiling experiments [15, 16]. The relative synonymous codon usage (RSCU) values for each codon in the reference set are then computed [17, 18]. For each codon in the target viral gene, the RSCU value from the reference set is assigned, and the geometric mean of these values across all codons in the gene yields the CAI [19, 20].
CAI values have been calculated for numerous veterinary viruses to assess host adaptation. For example, analysis of canine circovirus revealed CAI values that suggested moderate adaptation to canine hosts, with evidence of ongoing evolutionary optimization [3]. Similarly, studies of bovine ephemeral fever virus demonstrated CAI values consistent with adaptation to bovine hosts, although with notable variation among different genomic regions [10].
Effective Number of Codons (ENC)
The effective number of codons (ENC) provides a complementary measure of codon usage bias that is independent of any reference set [21, 22]. ENC quantifies the degree of codon usage bias within a single gene or genome, with values ranging from 20 (extreme bias, where only one codon is used per amino acid) to 61 (no bias, where all codons are used with equal frequency) [23, 24].
ENC values are particularly useful for comparing codon usage bias across different viral species or strains without requiring host-specific reference data [25, 26]. When combined with CAI analysis, ENC provides a comprehensive picture of both the overall bias within a viral genome and the specific adaptation to a particular host [27, 28].
Relative Synonymous Codon Usage (RSCU)
Relative synonymous codon usage (RSCU) values represent the observed frequency of a codon divided by its expected frequency under the assumption of equal usage among synonymous codons [29, 30]. RSCU values greater than 1 indicate overrepresented codons, while values less than 1 indicate underrepresented codons [31, 32].
RSCU analysis is typically performed on a codon-by-codon basis and can reveal specific patterns of codon preference that may be associated with host adaptation [33, 34]. For example, analysis of Lassa virus revealed distinct RSCU patterns for different genomic segments, suggesting segment-specific selective pressures [6]. Similarly, studies of ranavirus DNA polymerase genes identified conserved RSCU patterns across different viral isolates [7].
Dinucleotide Frequency Analysis
Dinucleotide frequency analysis examines the observed versus expected frequencies of all 16 possible dinucleotide combinations [35, 1]. The odds ratio for each dinucleotide is calculated as the ratio of observed to expected frequency, with values below 1 indicating underrepresentation and values above 1 indicating overrepresentation [2, 3].
CpG dinucleotide frequency is of particular interest in viral genomics due to its association with host immune recognition [4, 5]. Comprehensive analysis of poxvirus genomes revealed substantial variation in CpG content across different species, with some viruses exhibiting extreme CpG depletion [9]. This variation may reflect differences in host range, transmission mode, or immune evasion strategies [11, 12].
Host Adaptation and Evolutionary Dynamics
Natural Selection versus Mutational Pressure
The relative contributions of natural selection and mutational pressure to viral codon usage bias have been extensively debated [13, 14]. Mutational pressure refers to the tendency of viral polymerases to introduce specific nucleotide substitutions at characteristic rates, which can shape overall nucleotide composition and, consequently, codon usage [15, 16].
Neutrality plots, which compare the GC content at the third codon position (GC3) with the overall GC content, provide a method for distinguishing between these forces [17, 18]. A strong correlation between GC3 and overall GC content suggests that mutational pressure is the dominant force, while a weak correlation indicates that natural selection plays a more significant role [19, 20].
Studies of SARS-CoV-2 variants have demonstrated that natural selection plays a significant role in governing codon usage bias, particularly in variants of concern [17]. Similarly, analysis of coronaviruses from the Coronaviridae family revealed that natural selection shapes the codon usage of structural genes [28]. For African swine fever virus, both mutational pressure and natural selection contribute to observed codon usage patterns [15].
Host Switching and Codon Adaptation
When viruses cross species barriers and establish infection in new hosts, they face the challenge of adapting to a different tRNA pool and cellular environment [21, 22]. The degree of initial codon adaptation to the new host can influence the probability of successful host switching and subsequent viral emergence [23, 24].
Comparative analysis of monkeypox virus genomes from the 2022 outbreak revealed evidence of ongoing host adaptation at the codon level [11]. The study identified specific codons that showed significant differences in usage between human-derived and animal-derived isolates, suggesting selection for human-adapted variants [8, 11]. Similarly, analysis of hantaviruses causing hemorrhagic fever with renal syndrome demonstrated host-specific codon usage patterns that correlated with the rodent reservoir species [12].
Temporal Dynamics of Codon Usage Evolution
Viral codon usage patterns are not static but evolve over time in response to changing selective pressures [1, 2]. Longitudinal studies of SARS-CoV-2 genomes have documented temporal shifts in codon usage that correlate with the emergence of new variants [1, 2, 5]. These shifts may reflect ongoing adaptation to human hosts or the relaxation of selective constraints during rapid transmission [14, 18].
For veterinary viruses, temporal analysis of codon usage evolution can provide insights into the dynamics of host adaptation. Studies of canine circovirus have identified temporal trends in codon usage that correlate with geographic spread and host range expansion [3]. Similarly, analysis of bluetongue virus genomes from different time periods has revealed codon usage changes that may be associated with adaptation to new vector species or vertebrate hosts [33].
Applications in Veterinary Virology
Vaccine Development and Attenuation
Codon usage analysis has direct applications in veterinary vaccine development [13, 19]. Codon deoptimization, the process of introducing synonymous mutations that reduce codon adaptation to the host, can be used to generate live attenuated vaccines [13, 19]. By replacing optimal codons with rare codons, viral protein synthesis is reduced, leading to attenuated replication while maintaining immunogenicity [13, 19].
This approach has been successfully applied to multiple veterinary viruses. For example, codon-deoptimized variants of porcine reproductive and respiratory syndrome virus (PRRSV) have been generated and evaluated as vaccine candidates [13]. The deoptimized viruses showed reduced replication in cell culture and attenuated virulence in animal models while inducing protective immune responses [13].
Predicting Host Range and Zoonotic Potential
Bioinformatics analysis of codon usage bias can contribute to the assessment of viral host range and zoonotic potential [4, 11]. Machine learning approaches that incorporate codon usage features have been developed to predict viral host associations [4]. These methods can identify viruses that show unusual patterns of codon adaptation that may indicate recent host switching or the potential for cross-species transmission [4, 11].
For example, tree-based learning algorithms trained on codon usage biases and genomic characteristics have been used to predict viral host fitness and potential for host shifting [4]. These computational tools can assist in the surveillance of emerging veterinary pathogens and the identification of viruses with pandemic potential [4, 11].
Diagnostic Assay Design
Codon usage analysis can inform the design of molecular diagnostic assays for veterinary viruses [21, 22]. Regions of the viral genome that show high codon usage bias may be under strong selective constraint and therefore more conserved across viral strains [21, 22]. Targeting these regions for primer or probe design can improve assay sensitivity and specificity [21, 22].
Conversely, regions with low codon usage bias may be more variable and suitable for strain typing or phylogenetic analysis [23, 24]. Understanding the codon usage landscape of a viral genome can therefore guide the selection of optimal target regions for different diagnostic applications [23, 24].
Workflow for Codon Usage Analysis
The following Mermaid diagram illustrates a typical bioinformatics workflow for analyzing viral codon usage bias and host adaptation:
flowchart TD
A[Viral Genome Sequences], > B[Quality Control and Assembly]
B, > C[Codon Extraction and Counting]
C, > D[Calculate RSCU Values]
D, > E[Compute ENC Values]
E, > F[Calculate CAI against Host Reference]
F, > G[Dinucleotide Frequency Analysis]
G, > H[Neutrality Plot Analysis]
H, > I[Correlation with Host tRNA Abundance]
I, > J[Phylogenetic Comparative Analysis]
J, > K[Interpretation: Host Adaptation Assessment]
K, > L[Vaccine Design Applications]
K, > M[Host Range Prediction]
K, > N[Diagnostic Target Selection]
Structural Constraints and mRNA Folding
The relationship between codon usage and mRNA secondary structure represents an important but often overlooked aspect of viral genome evolution [24, 25]. Computational RNA folding algorithms, such as those based on thermodynamic minimization, can predict the secondary structure of viral mRNAs and identify regions where codon choice is constrained by structural requirements [24, 25].
Color-coded 3D representations of mRNA secondary structure provide a visual tool for understanding these constraints. In these representations, each nucleotide is positioned in three-dimensional space according to the predicted folding pattern, and the color of each nucleotide reflects its thermodynamic contribution to the overall structure [27, 28]. Nucleotides colored in blue typically represent regions of high stability that are deeply embedded in stem structures, while nucleotides colored in red represent regions of low stability that are located in loops or single-stranded regions [27, 28].
Studies of intrinsically disordered protein regions in viral genomes have revealed that these regions often exhibit distinct codon usage patterns compared to ordered regions [24]. The structural flexibility required for intrinsically disordered proteins may impose specific constraints on mRNA folding that influence codon choice [24]. This relationship between protein structure, mRNA folding, and codon usage represents an active area of research in viral bioinformatics [24, 25].
Comparative Analysis Across Viral Families
DNA Viruses
DNA viruses, including poxviruses, circoviruses, and ranaviruses, exhibit diverse codon usage patterns that reflect their different replication strategies and host ranges [7, 8, 9]. Poxviruses, which replicate in the cytoplasm using their own DNA-dependent RNA polymerase, show distinct nucleotide and dinucleotide composition patterns compared to nuclear-replicating DNA viruses [9]. Analysis of monkeypox virus genomes has revealed codon usage patterns that are intermediate between those of their natural rodent hosts and their human hosts, suggesting ongoing adaptation [8, 11].
Circoviruses, which are among the smallest known viruses, show extreme codon usage bias that may reflect the severe constraints imposed by their compact genomes [3, 19]. Studies of canine circovirus have identified specific codons that are consistently overrepresented or underrepresented across global isolates, suggesting strong selective pressures [3]. Similarly, analysis of porcine circoviruses has revealed host-specific codon usage patterns that correlate with pathogenicity [19].
RNA Viruses
RNA viruses, including coronaviruses, flaviviruses, and orthomyxoviruses, show codon usage patterns that are strongly influenced by the high mutation rates of their RNA-dependent RNA polymerases [1, 2, 30]. The biased mutation spectra of these polymerases can drive the overall nucleotide composition of viral genomes, which in turn affects codon usage [30].
Coronaviruses have been extensively studied from a codon usage perspective due to their large RNA genomes and pandemic potential [1, 2, 5, 13, 14, 17, 18, 22, 23, 25, 28, 31, 32, 35]. These studies have revealed that coronavirus codon usage is shaped by both mutational pressure and natural selection, with different genomic regions showing different degrees of host adaptation [1, 2, 28]. The spike protein gene, for example, often shows distinct codon usage patterns compared to other structural and nonstructural genes [13, 17].
Bluetongue virus, a segmented RNA virus that infects ruminants, has been analyzed for codon usage bias across its 10 genomic segments [33]. The study revealed segment-specific codon usage patterns that may reflect different functional constraints on the encoded proteins [33]. Similarly, analysis of Crimean-Congo hemorrhagic fever virus segments showed distinct codon usage patterns for the S, M, and L segments [21].
Retroviruses and Reverse-Transcribing Viruses
Retroviruses and other reverse-transcribing viruses face unique constraints on codon usage due to the dual requirement for both RNA and DNA replication intermediates [27, 34]. The codon usage of these viruses is influenced by both the host tRNA pool and the specific requirements of reverse transcription [27, 34].
Analysis of anelloviruses, which are circular single-stranded DNA viruses that replicate through a rolling-circle mechanism, has revealed complex codon usage patterns that vary among different viral species [27]. The study identified both mutational and selective forces shaping codon usage in this diverse viral family [27].
Limitations and Considerations
Several methodological considerations must be addressed when interpreting codon usage analyses. First, the choice of reference set for CAI calculation can substantially influence the results [9, 10]. Different tissues within a host may have different tRNA profiles, and the optimal reference set should reflect the specific cell types that are targeted by the virus [9, 10].
Second, codon usage bias can be influenced by factors other than host adaptation, including genomic base composition, replication strategy, and the presence of RNA secondary structures [24, 25]. These confounding factors must be carefully considered when drawing conclusions about host adaptation from codon usage data [24, 25].
Third, the relationship between codon usage and translational efficiency is not always straightforward [11, 12]. While codon optimization generally correlates with increased protein production, other factors such as mRNA structure, codon context, and the availability of specific tRNAs can modulate this relationship [11, 12].
Future Directions
The field of viral codon usage analysis continues to evolve with the development of new computational methods and the accumulation of genomic sequence data. Machine learning approaches that integrate codon usage features with other genomic characteristics are being developed to improve predictions of viral host range and emergence potential [4].
Single-cell transcriptomics and tRNA sequencing technologies are providing unprecedented resolution into the cell-type-specific tRNA pools that viruses encounter during infection [9, 10]. These data will enable more accurate assessments of codon adaptation at the cellular level rather than the organismal level [9, 10].
The integration of codon usage analysis with structural biology approaches, including cryo-electron microscopy and computational protein structure prediction, promises to reveal new insights into the relationship between codon choice, mRNA folding, and protein function [24, 25].
Conclusion
Bioinformatics analysis of viral codon usage bias provides a powerful framework for understanding host adaptation, evolutionary dynamics, and translational regulation in veterinary viruses. The combination of metrics such as CAI, ENC, and RSCU with dinucleotide frequency analysis and mRNA folding predictions enables comprehensive characterization of the selective pressures acting on viral genomes. These analyses have direct applications in vaccine development, host range prediction, and diagnostic assay design. As genomic sequencing technologies continue to advance and computational methods become more sophisticated, codon usage analysis will remain an essential tool in the veterinary virologist's arsenal for understanding and controlling viral diseases.
References
[1] Błażej P, Mackiewicz D, Mackiewicz P. Temporal and gene-specific dynamics of codon usage evolution in SARS-CoV-2 genomes. BMC Genomics. 2026. https://pubmed.ncbi.nlm.nih.gov/41735839/
[2] Davidson A, Parr M, Totzeck F, et al. Over time analysis of the codon usage of SARS-CoV-2 and its variants. Comput Struct Biotechnol J. 2025. https://pubmed.ncbi.nlm.nih.gov/40502937/
[3] Li G, Liu R, Chen Y, et al. Evolutionary dynamics and codon bias analysis of canine circovirus: Insights into global spread and host adaptability. Mol Phylogenet Evol. 2025. https://pubmed.ncbi.nlm.nih.gov/40335001/
[4] Su S, Ni Z, Lan T, et al. Predicting viral host codon fitness and path shifting through tree-based learning on codon usage biases and genomic characteristics. Sci Rep. 2025. https://pubmed.ncbi.nlm.nih.gov/40211017/
[5] Padhiar NH, Ghazanchyan T, Fumagalli SE, et al. SARS-CoV-2 CoCoPUTs: analyzing GISAID and NCBI data to obtain codon statistics, mutations, and free energy over a multiyear period. Virus Evol. 2025. https://pubmed.ncbi.nlm.nih.gov/39882309/
[6] Rahman SU, Hu Y, Rehman HU, et al. Analysis of synonymous codon usage bias of Lassa virus. Virus Res. 2025. https://pubmed.ncbi.nlm.nih.gov/39832535/
[7] Aktürk Dizman Y. Exploring Codon Usage Patterns and Influencing Factors in Ranavirus DNA Polymerase Genes. J Basic Microbiol. 2024. https://pubmed.ncbi.nlm.nih.gov/39099168/
[8] Shan KJ, Wu C, Tang X, et al. Molecular Evolution of Protein Sequences and Codon Usage in Monkeypox Viruses. Genomics Proteomics Bioinformatics. 2024. https://pubmed.ncbi.nlm.nih.gov/38862422/
[9] Molteni C, Forni D, Cagliani R, et al. Evolution and diversity of nucleotide and dinucleotide composition in poxviruses. J Gen Virol. 2023. https://pubmed.ncbi.nlm.nih.gov/37792576/
[10] Rani S, Mamathashree MN, Bharthi IU, et al. Comprehensive examination on codon usage bias pattern of the Bovine Ephemeral fever virus. J Biomol Struct Dyn. 2024. https://pubmed.ncbi.nlm.nih.gov/37705249/
[11] Zhou J, Wang X, Zhou Z, et al. Insights into the Evolution and Host Adaptation of the Monkeypox Virus from a Codon Usage Perspective: Focus on the Ongoing 2022 Outbreak. Int J Mol Sci. 2023. https://pubmed.ncbi.nlm.nih.gov/37511283/
[12] Noor F, Ashfaq UA, Bakar A, et al. Identification and characterization of codon usage pattern and influencing factors in HFRS-causing hantaviruses. Front Immunol. 2023. https://pubmed.ncbi.nlm.nih.gov/37492567/
[13] Wu X, Shan KJ, Zan F, et al. Optimization and Deoptimization of Codons in SARS-CoV-2 and Related Implications for Vaccine Development. Adv Sci (Weinh). 2023. https://pubmed.ncbi.nlm.nih.gov/37267926/
[14] Fumagalli SE, Padhiar NH, Meyer D, et al. Analysis of 3.5 million SARS-CoV-2 sequences reveals unique mutational trends with consistent nucleotide and codon frequencies. Virol J. 2023. https://pubmed.ncbi.nlm.nih.gov/36812119/
[15] Pu F, Wang R, Yang X, et al. Nucleotide and codon usage biases involved in the evolution of African swine fever virus: A comparative genomics analysis. J Basic Microbiol. 2023. https://pubmed.ncbi.nlm.nih.gov/36782108/
[16] Shueb M, Prasad SK, Suresh KP, et al. The first study on analysis of the codon usage bias and evolutionary analysis of the glycoprotein envelope E2 gene of seven Pestiviruses. Vet World. 2022. https://pubmed.ncbi.nlm.nih.gov/36185504/
[17] Tyagi N, Sardar R, Gupta D. Natural selection plays a significant role in governing the codon usage bias in the novel SARS-CoV-2 variants of concern (VOC). PeerJ. 2022. https://pubmed.ncbi.nlm.nih.gov/35765592/
[18] Mogro EG, Bottero D, Lozano MJ. Analysis of SARS-CoV-2 synonymous codon usage evolution throughout the COVID-19 pandemic. Virology. 2022. https://pubmed.ncbi.nlm.nih.gov/35134624/
[19] Feng H, Segalés J, Wang F, et al. Comprehensive Analysis of Codon Usage Patterns in Chinese Porcine Circoviruses Based on Their Major Protein-Coding Sequences. Viruses. 2022. https://pubmed.ncbi.nlm.nih.gov/35062285/
[20] Rahman SU, Abdullah M, Khan AW, et al. A detailed comparative analysis of codon usage bias in Alongshan virus. Virus Res. 2022. https://pubmed.ncbi.nlm.nih.gov/34822954/
[21] Beelagi MS, Kumar SS, Indrabalan UB, et al. Synonymous codon usage pattern among the S, M, and L segments in Crimean-congo hemorrhagic fever virus. Bioinformation. 2021. https://pubmed.ncbi.nlm.nih.gov/34602775/
[22] Khattak S, Rauf MA, Zaman Q, et al. Genome-Wide Analysis of Codon Usage Patterns of SARS-CoV-2 Virus Reveals Global Heterogeneity of COVID-19. Biomolecules. 2021. https://pubmed.ncbi.nlm.nih.gov/34207362/
[23] Das JK, Roy S. Comparative analysis of human coronaviruses focusing on nucleotide variability and synonymous codon usage patterns. Genomics. 2021. https://pubmed.ncbi.nlm.nih.gov/34019999/
[24] Kumar N, Kaushik R, Tennakoon C, et al. Insights into the evolutionary forces that shape the codon usage in the viral genome segments encoding intrinsically disordered protein regions. Brief Bioinform. 2021. https://pubmed.ncbi.nlm.nih.gov/33866372/
[25] Saha J, Bhattacharjee S, Pal Sarkar M, et al. A comparative genomics-based study of positive strand RNA viruses emphasizing on SARS-CoV-2 utilizing dinucleotide signature, codon usage and codon context analyses. Gene Rep. 2021. https://pubmed.ncbi.nlm.nih.gov/33615042/
[26] Nguyen TH, Wang D, Rahman SU, et al. Analysis of codon usage patterns and influencing factors in rice tungro bacilliform virus. Infect Genet Evol. 2021. https://pubmed.ncbi.nlm.nih.gov/33548490/
[27] Deb B, Uddin A, Chakraborty S. Composition, codon usage pattern, protein properties, and influencing factors in the genomes of members of the family Anelloviridae. Arch Virol. 2021. https://pubmed.ncbi.nlm.nih.gov/33392821/
[28] Nyayanit DA, Yadav PD, Kharde R, et al. Natural Selection Plays an Important Role in Shaping the Codon Usage of Structural Genes of the Viruses Belonging to the Coronaviridae Family. Viruses. 2020. https://pubmed.ncbi.nlm.nih.gov/33375017/
[29] Faye M, Zein N, Loucoubar C, et al. Biological Characteristics and Patterns of Codon Usage Evolution for the African Genotype Zika Virus. Viruses. 2020. https://pubmed.ncbi.nlm.nih.gov/33202554/
[30] Kustin T, Stern A. Biased Mutation and Selection in RNA Viruses. Mol Biol Evol. 2021. https://pubmed.ncbi.nlm.nih.gov/32986832/