SEARCH · PubMed Health

Results for “Codon”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 505 records · Page 28Linked to original sources

PubMed2000

Nucleotides of 18S rRNA surrounding mRNA codons at the human ribosomal A, P, and E sites: a crosslinking study with mRNA analogs carrying an aryl azide group at either the uracil or the guanine residue.

The 18S rRNA environment of the mRNA at the decoding site of human 80S ribosomes has been studied by cross-linking with derivatives of hexaribonucleotide UUUGUU (comprising Phe and Val codons) that carried a perfluorophenylazide group either at the N7 atom of the guanine or at the C5 atom of the 5'-terminal uracil residue. The location of the codons on the ribosome at A, P, or E sites has been adjusted by the cognate tRNAs. Three types of complexes have been obtained for each type derivative, namely, (1) codon UUU and Phe-tRNAPhe at the P site (codon GUU at the A site), (2) codon UUU and tRNAPhe at the P site and PheVal-tRNAVal at the A site, and (3) codon GUU and Val-tRNAVal at the P site (codon UUU at the E site). This allowed the placement of modified nucleotides of the mRNA analog at positions -3, +1, or +4 on the ribosome. Mild UV irradiation resulted in tRNA-dependent crosslinking of the mRNA analogs to the 18S rRNA. Nucleotide G961 crosslinked to mRNA position -3, nucleotide G1207 to position +1, and A1823 together with A1824 to position +4. All of these nucleotides are located in the most strongly conserved regions of the small subunit RNA structure, and correspond to nucleotides G693, G926, G1491, and A1492 of bacterial 16S rRNA. Three of them (with the exception of G1491) had been found earlier at the 70S ribosomal decoding site. The similarities and differences between the 16S and 18S rRNA decoding sites are discussed.

Anticodon↗

PubMed1997

Codon 201Arg/Gly polymorphism of DCC (deleted in colorectal carcinoma) gene in flat- and polypoid-type colorectal tumors.

Recent studies have identified the distinct existence of flat-type colorectal tumors. The low incidence of ras gene mutations in these tumors suggests that their genetic pathways of tumor progression may be different from those of the polypoid type. To elucidate further genetic alterations in flat-type colorectal tumors, codon 201Arg/Gly polymorphism in the DCC (deleted in colorectal carcinoma) gene was analyzed in normal tissue (normal colonic mucosa or peripheral lymphocytes) and in tumor tissue from 191 patients with colorectal tumors (36 patients with flat-type colorectal tumors, 81 patients with polypoid-type colorectal tumors, and 74 patients with advanced carcinomas). For normal controls, 30 samples obtained from patients who had neither colorectal tumors (confirmed by total colonoscopy) nor a family history of colorectal carcinoma were analyzed. DCC gene codon 201Arg/Gly polymorphism was investigated by polymerase chain reaction-based restriction fragment length polymorphism analysis, fluorescence-based dideoxy sequencing, or both. For the flat type, the frequency of codon 201Gly of the DCC gene was 64% and 54% in the normal tissue of patients with adenoma with high-grade dysplasia and submucosal carcinoma, respectively. It was 49%, 52%, and 49% in the normal tissue of patients with polypoid-type adenoma with high-grade dysplasia, submucosal carcinoma, and advanced carcinoma, respectively. In the normal tissue, codon 201Gly of the DCC gene was more frequently observed in patients with flat-type adenoma with low-grade dysplasia (67%) than in those with polypoid-type adenoma with low-grade dysplasia (18%) or in normal controls (17%, P < 0.05, chi2 test). Codon 201Arg/Gly polymorphism in tumor tissues did not differ from that in the corresponding normal tissues, except for 10 cases of carcinoma with loss of heterozygosity (LOH). In carcinomas with LOH, preferential loss of the codon 201Arg allele was noted (9/10 cases). These results suggest that codon 201Gly of the DCC gene is not only associated with flat-type colorectal tumors, but that it may serve as a useful genetic marker for identifying groups at higher risk for colorectal cancer.

Carcinoma↗

PubMed2004

Nucleotides downstream of start codons show marked non-randomness in Escherichia coli but not in Bacillus subtilis.

This study aimed at measuring the nucleotide non-randomness in the region downstream of start codons in bacterial genes and to see if the non-randomness differs between biased and unbiased genes, in terms of the effective number of codons (Nc) and the codon adaptation index (CAI). In Escherichia coli, there was a marked elevation in nucleotide conservation for the genes having low Nc-values compared to the genes having high Nc-values, i.e the more biased genes showed a higher level of non-randomness. Likewise, the genes displaying high CAI-values showed stronger nucleotide conservation than the genes of low CAI-values. This elevated conservation is visible up to approximately 15-17 nucleotides downstream of the start codon, after which there is little difference. This indicates that there may be distinct selectional mechanisms acting upon the first 5-6 codons within genes in E. coli. In B. subtilis, these effects are less pronounced, if present at all. Furthermore, analyses of codons used in this region were not in support of the hypothesis that the elevation in nucleotide non-randomness is a question of selection for certain optimal codons.

Bacillus subtilis↗

PubMed1998

Detection of the DCC gene product in normal and malignant colorectal tissues and its relation to a codon 201 mutation.

Protein expression of the putative tumour-suppressor gene DCC on chromosome 18q was evaluated in a panel of 16 matched colorectal cancer and normal colonic tissue samples together with DCC mRNA expression and allelic deletions (loss of heterozygosity, LOH). Determined by a polymerase chain reaction (PCR)-LOH assay, 12 of the 16 (75%) cases were informative with LOH occurring in 2 of the 12 cases. For DCC mRNA, transcripts could be detected in all analysed normal tissues (eight out of eight) by RT-PCR, whereas 6 of the 15 tumours were negative. DCC protein expression, investigated by immunohistochemistry using the monoclonal antibody 15041 A directed against the intracellular domain, was homogeneously positive in all normal tissue samples. In tumour tissues, no DCC protein was seen in 11 out of 16 samples (69%). For the DCC codon 201, we found a loss of a wild-type codon sequence caused by mutation or LOH in at least 8 out of 15 cases (53%) compared with the corresponding normal tissue. DCC protein expression was undetectable in eight of the nine tumours missing both wild-type codons. Only one of the five tumours with retained DCC protein expression had no detectable wild-type codon 201. In addition, 9 out of 15 normal tissue specimens were mutated in codon 201. In two out of three cases with homozygous wild-type codons in peripheral blood lymphocyte (PBL) DNA, mutations were already observed in the tumour adjacent normal colonic mucosa. We conclude that DCC immunostaining should be introduced in the clinicopathological routine because of its strong correlation with the known prognostic markers 18q LOH and mutation of codon 201.

Blotting, Southern↗

PubMed2016

Expanding the amino acid repertoire of ribosomal polypeptide synthesis via the artificial division of codon boxes.

In ribosomal polypeptide synthesis the library of amino acid building blocks is limited by the manner in which codons are used. Of the proteinogenic amino acids, 18 are coded for by multiple codons and therefore many of the 61 sense codons can be considered redundant. Here we report a method to reduce the redundancy of codons by artificially dividing codon boxes to create vacant codons that can then be reassigned to non-proteinogenic amino acids and thereby expand the library of genetically encoded amino acids. To achieve this, we reconstituted a cell-free translation system with 32 in vitro transcripts of transfer RNASNN (tRNASNN) (S = G or C), assigning the initiator and 20 elongator amino acids. Reassignment of three redundant codons was achieved by replacing redundant tRNASNNs with tRNASNNs pre-charged with non-proteinogenic amino acids. As a demonstration, we expressed a 32-mer linear peptide that consists of 20 proteinogenic and three non-proteinogenic amino acids, and a 14-mer macrocyclic peptide that contains more than four non-proteinogenic amino acids.

Amino Acid Sequence↗

PubMed2005

Detection of point mutations at codon 12 of KI-ras in ophthalmic pterygia.

AIMS: Ophthalmic pterygium is a potentially vision-threatening lesion of unknown etiology, related to an exposure to solar light. Mutations to the ras genes are frequently observed in lesions related to an exposure to solar light. The present study aims at screening pterygia for mutations at codons 12 and 13 of the ras genes. METHODS: In all, 50 pterygia were examined, together with respective blood samples and specimens of normal conjunctiva. A PCR reaction was performed to amplify sequences containing codons 12 and 13 of Ki-ras, H-ras, and N-ras. An RFLP analysis was then performed to detect point mutations at codon 12. The mutational status at codons 12 and 13 was further explored with sequencing of PCR products. RESULTS: RFLP analysis revealed Ki-ras mutations at codon 12 in five (10%) of pterygia, whereas H-ras or N-ras mutations were not observed. Sequencing confirmed Ki-ras mutations at codon 12 and revealed absence of mutations at codon 13. The presence of Ki-ras mutations was significantly correlated with postoperative recurrence (P=0.02) and young age (P=0.04). Mutations were not observed in specimens of blood or normal conjunctiva for any of the genes examined. CONCLUSIONS: The absence of N-ras mutations is in agreement with previous reports concerning mucosal lesions. The detection of Ki-ras mutations and the association with postoperative recurrence implies a possible role of Ki-ras in the clinical profile of pterygium. The mechanism of Ki-ras mutations is unclear and could be independent of the action of UV light.

Adult↗

PubMed2001

Does recombination improve selection on codon usage? Lessons from nematode and fly complete genomes.

Understanding the factors responsible for variations in mutation patterns and selection efficacy along chromosomes is a prerequisite for deciphering genome sequences. Population genetics models predict a positive correlation between the efficacy of selection at a given locus and the local rate of recombination because of Hill-Robertson effects. Codon usage is considered one of the most striking examples that support this prediction at the molecular level. In a wide range of species including Caenorhabditis elegans and Drosophila melanogaster, codon usage is essentially shaped by selection acting for translational efficiency. Codon usage bias correlates positively with recombination rate in Drosophila, apparently supporting the hypothesis that selection on codon usage is improved by recombination. Here we present an exhaustive analysis of codon usage in C. elegans and D. melanogaster complete genomes. We show that in both genomes there is a positive correlation between recombination rate and the frequency of optimal codons. However, we demonstrate that in both species, this effect is due to a mutational bias toward G and C bases in regions of high recombination rate, possibly as a direct consequence of the recombination process. The correlation between codon usage bias and recombination rate in these species appears to be essentially determined by recombination-dependent mutational patterns, rather than selective effects. This result highlights that it is necessary to take into account the mutagenic effect of recombination to understand the evolutionary role and impact of recombination.

Animals↗

PubMed1975

Allosteric mechanism for codon-dependent tRNA selection on ribosomes.

We suggest that the interaction between a codon and its cognate tRNA induces conformational changes in the tRNA. We further suggest that sites on the ribosome preferentially bind tRNA in those conformations which require proper matching of codon and anticodon. According to this model, the codon functions as an allosteric effector which influences the conformation at various sites in the tRNA. This is made possible by the ribosome, which we suggest traps tRNA molecules in those conformation states that maximize the energy difference between cognate and noncognate codon-anticodon interactions. Studies of the interactions between tRNA molecules and their cognate codons in the absence of the ribosome have suggested that triplet-triplet interaction between codon and anticodon is far too weak to account for the specificity of the tRNA selection mechanism during protein synthesis. In contrast, we suggest that such affinity measurements do not adequately describe the interaction between a codon and its cognate tRNA. Thus, such experiments can not detect conformational changes in the tRNA, and, in particular, those stabilized by the ribosome.

Allosteric Regulation↗

PubMed1985

Direct mapping of adeno-associated virus capsid proteins B and C: a possible ACG initiation codon.

The three major capsid proteins of adeno-associated virus type 2 (AAV2) virions are designated A, B, and C and have molecular sizes of 90, 72, and 60 kDa, respectively. These proteins are related, and genetic studies have shown they are encoded by a long open reading frame located in the right half of the genome. The coding capacity distal to the first ATG in this reading frame is only 503 amino acids (i.e., a protein about the size of protein C), but an open frame sequence devoid of ATG codons extends upstream for an additional 184 codons. Although the amino terminus of the C capsid protein is blocked, partial amino acid sequence analyses of peptides from C have confirmed that it is encoded within the portion of the reading frame distal to the first ATG at nucleotide (nt) location 2810. The amino terminus of the B capsid protein is not blocked, and its sequence begins with alanine. The triplet encoding this alanine lies 64 codons upstream from the initiation site for protein C and is immediately preceded by the threonine codon, ACG, at nt 2615. This ACG codon lies in the most favorable sequence context for protein synthesis initiation. All three AAV2 capsid proteins are labeled in vitro with formyl[35S]methionyl-tRNAf, indicating that synthesis of each protein is initiated independently. Our data suggest that the nt 2615 ACG codon directs the methionyl-tRNA-dependent initiation of the AAV2 B capsid protein. Proteins B and C may be synthesized from the same mRNA species and their relative abundance could be determined by the efficiencies of their respective initiation codons.

Amino Acid Sequence↗

PubMed1986

Transcription attenuation in Salmonella typhimurium: the significance of rare leucine codons in the leu leader.

The leucine operon of Salmonella typhimurium is controlled by a transcription attenuation mechanism. Four adjacent leucine codons within a 160-nucleotide leu leader RNA are thought to play a central role in this mechanism. Three of the four codons are CUA, a rarely used leucine codon within enteric bacteria. To determine whether the nature of the leucine codon affects the regulation of the leucine operon, we used oligonucleotide-directed mutagenesis to first convert one CUA of the leader to CUG and then convert all three CUA codons to CUG. CUG is the most frequently used leucine codon in enteric bacteria. A mutant having (CUA)2CUGCUC in place of (CUA)3CUC has an altered response to leucine limitation, requiring a slightly higher degree of limitation to effect derepression. Changing (CUA)3CUC to (CUG)3CUC has more dramatic effects upon operon expression. First, the basal level of expression is lowered to the point that the mutant grows more slowly than the parent in a minimal medium lacking leucine. Second, the response of the mutant to a leucine limitation is dramatically altered such that even a strong limitation elicits only a modest degree of derepression. If the mutant is grown under conditions of leucyl-tRNA limitation rather than leucine limitation, complete derepression can be achieved, but only at a much higher degree of limitation than for the wild-type operon. These results provide a clear-cut example of codon usage having a dramatic effect upon gene expression.

Codon↗

PubMed1989

Nonrandom utilization of codon pairs in Escherichia coli.

We have analyzed protein-coding sequences of Escherichia coli and find that codon-pair utilization is highly biased, reflecting overrepresentation or underrepresentation of many pairs compared with their random expectations. This effect is over and above that contributed by nonrandomness in the use of amino acid pairs, which itself is highly evident; it is much weaker when nonadjacent codon pairs are examined and virtually disappears when pairs separated by two or three intervening codons are evaluated. There appears to be a high degree of directionality in this bias: any codon that participates in many nonrandom pairs tends to make both over- and underrepresented pairs, but preferentially as a left- or right-hand member. We show a relationship between codon-pair utilization patterns and levels of gene expression: genes encoding proteins expressed at high levels tend to contain more abundant, but more highly underrepresented, codon pairs, relative to genes expressed at low levels. The nonrandom utilization of codon pairs may be a consequence of their effects on translational efficiency, which in turn may be related to the compatibility of adjacent aminoacyl-tRNA isoacceptors at the A and P sites of a translating ribosome.

Bacterial Proteins↗

PubMed1989

Codon discrimination and anticodon structural context.

Site-directed mutagenesis has been used to change the nucleotide C in the wobble position of tRNA(1Gly) (CCC) to U. The mutated tRNA was tested for its ability to read glycine codons in an in vitro protein-synthesizing system programmed with the phage message MS2-RNA that had been modified by site-directed mutagenesis so as to make it possible to monitor conveniently the reading of all four glycine codons. The results showed that while the efficiency of tRNA(1Gly) (UCC) was comparable to that of mycoplasma tRNA(Gly) (UCC) in the reading of the codon GGA, the mycoplasma tRNA(Gly) was far more efficient than the tRNA(1Gly) (UCC) in the reading of the codons GGU and GGC. Thus, the anticodon UCC, when present in the structural context of the tRNA(1Gly) molecule, behaved as predicted by the wobble rules while in the structural context of the mycoplasma tRNA(Gly) it read without discrimination between the nucleotides in the third codon position, in violation of the wobble restrictions. The result with the codon GGG showed that the anticodon UCC, when present in tRNA(1Gly), was considerably less efficient in reading this codon than it was in the structural context of the mycoplasma tRNA(Gly). It would therefore seem that the anticodon UCC, when present in a certain tRNA, can be an efficient wobbler, while in the molecular environment of another tRNA it is markedly restricted in its ability to wobble.

Amino Acid Sequence↗

PubMed1990

Features of the formate dehydrogenase mRNA necessary for decoding of the UGA codon as selenocysteine.

The fdhF gene encoding the 80-kDa selenopolypeptide subunit of formate dehydrogenase H from Escherichia coli contains an in-frame TGA codon at amino acid position 140, which encodes selenocysteine. We have analyzed how this UGA "sense codon" is discriminated from a UGA codon signaling polypeptide chain termination. Deletions were introduced from the 3' side into the fdhF gene and the truncated 5' segments were fused in-frame to the lacZ reporter gene. Efficient read-through of the UGA codon, as measured by beta-galactosidase activity and incorporation of selenium, was dependent on the presence of at least 40 bases of fdhF mRNA downstream of the UGA codon. There was excellent correlation between the results of the deletion studies and the existence of a putative stem-loop structure lying immediately downstream of the UGA in that deletions extending into the helix drastically reduced UGA translation. Similar secondary structures can be formed in the mRNAs coding for other selenoproteins. Selenocysteine insertion cartridges were synthesized that contained this hairpin structure and variable portions of the fdhF gene upstream of the UGA codon and inserted into the lacZ gene. Expression studies showed that upstream sequences were not required for selenocysteine insertion but that they may be involved in modulating the efficiency of read-through. Translation of the UGA codon was found to occur with high fidelity since it was refractory to ribosomal mutations affecting proofreading and to suppression by the sup-9 gene product.

Aldehyde Oxidoreductases↗

PubMed1995

Ribosomal decoding processes at codons in the A or P sites depend differently on 2'-OH groups.

The importance of 2'-OH groups of codons for binding of cognate tRNAs to ribosomal P and A sites was analyzed applying the following strategy. An mRNA of 41 nucleotides was synthesized with the structure C16-GAA-UUC-GUC-C16 coding for glutamic acid (E), phenylalanine (F) and valine (V), respectively, in the middle (EFV-mRNA). A second template, the E(dF)V-mRNA, was identical except that it carried a deoxyribo-codon-dUdUdC- for phenylalanine. tRNA binding to the P site is totally insensitive to the presence or absence of the 2'-OH group of the P-site codon, and tRNA binding to the P site is also not affected if the A-site codon lacks the 2'-OH groups. However, binding is impaired if the deoxy-codon is present at the E site. In sharp contrast, the A-site binding of Ac-aminoacyl-tRNA was severely reduced in the presence of the deoxy-codon at the A site as well as at the P site. The results demonstrate that the correctness of base pairing is also "sensed" via a correct sugar structure of the codon, e.g. positioning of the sugar pucker (2'-OH), during the decoding process at the A site (elongation) but not during the decoding at the P site (initiation).

Base Sequence↗

PubMed1984

Analysis of the codon bias in E. coli sequences.

Fifty-three gene sequences from E. coli containing 18,288 reading frame triplets have been characterized according to the nature and level of average codon preference. The distribution of average preferences is bimodal, with approximately half the genes using an average of only 36 codons, and the remainder just 42 codons. There is a high correlation between the level of codon bias, the tRNA population and the abundance of protein product, indicating biased patterns are exploited by the cell for the production of widely different levels of gene product. This relationship is especially striking in genes involved in the production of components for transcription and translation. Overall, the genes for these processes generate some five-fold more protein than the average in the genome, and use about five fewer codons. The very high codon bias found in the RNA polymerase gene thus provides a simple, autogenous mechanism for the coordinate synthesis of these components and RNA polymerase. A surprisingly high level of codon probability is also found in triplets of the complement of coding sequences. This is apparently due to the evolutionary dispersion of coding sequences and/or the requirement for increased levels of secondary structure in messenger RNAs.

Base Sequence↗

PubMed2004

RET proto-oncogene mutations are restricted to codon 634 and 618 in Korean families with multiple endocrine neoplasia 2A.

Identification of the germline mutation in the RET proto-oncogene is important for the diagnosis of hereditary medullary thyroid carcinoma (MTC). Hereditary forms account for approximately 25%-30% of all cases of MTC. The objective of this study was to evaluate the prevalence of the RET mutation and the genotype-phenotype relation in Korean patients with MTC. Genomic DNAs were obtained from 33 patients with MTC (M:F = 10:23, 39.8 +/- 12.0 years) who underwent total thyroidectomy between 1997 and 2003 at the Samsung Medical Center. Exons 10, 11, 13, 14, 15 and 16 of the RET proto-oncogene were amplified with specific primers using polymerase chain reaction (PCR). Sequence analysis was performed on the polymerase chain reaction (PCR) product using an automatic sequence analyzer. Nine of the 33 patients (M:F = 3:6, 33.3 +/- 10.0 years) were identified as having RET mutations. Six patients had multiple endocrine neoplasia (MEN) 2A and one had familial medullary thyroid carcinoma (FMTC). The remaining two patients were thought to have sporadic MTC. Five of the patients with MEN 2A had RET mutations in codon 634 of exon 11 (3 patients, C634Y; 2 patients, C634R) and the other patient with MEN 2A had a RET mutation in codon 618 of exon 10 (C618R). The patient with FMTC had a mutation in codon 634 (C634W). The two patients with sporadic MTC had RET mutations in codon 634 (1 patient, C634Y; 1 patient, C634S). We were not able to identify any genotype-phenotype relations because of the limited number of patients. Twenty-seven percent (9/33) of the patients with MTC in this study had RET mutations. Taking other studies into account, 77% (10/13) of Korean families with MEN 2A, including 7 other families in three reports from Korea, had RET mutations in codon 634 (5 families, C634Y; 4 families, C634R; 1 family, C634W), and 23% (3/13) had RET mutations in codon 618 (2 families, C618R; 1 family, C618S). RET proto-oncogene mutations were restricted to codon 634 and 618 in Korean families with MEN 2A.

Adult↗

PubMed2004

A stable upstream stem-loop structure enhances selection of the first 5'-ORF-AUG as a main start codon for translation initiation of human ACAT1 mRNA.

Human ACAT 1 cDNA K1 was first cloned and functionally expressed in 1993. There are two adjacent in-frame AUG codons, AUG(1397-1399) and AUG(1415-1417), at 5'-terminus of the open reading frame (ORF, nt 1397-3049) of human ACAT1 mRNA corresponding to cDNA K1. In current work, these two adjacent inframe AUGs at 5'-terminus of the predicted ORF (5'-ORF-AUGs) as start codons for translation initiation of human ACAT1 mRNA were characterized in detail. Codon mutations indicated that both of these two adjacent 5'-ORF-AUGs can be selected as start codons but the first 5'-ORF-AUG(1397-1399) is a main start codon consistent with that of the predicted ORF of human ACAT1 mRNA. Further deletion and mutation analyses demonstrated that a stable upstream stem-loop structure enhanced the selection of the first 5'-ORF-AUG(1397 -1399) as a main start codon, in addition to upstream nucleotide A in the -3 position, which is a key site of Kozak sequence. In addition, result of ACAT1 enzymatic activity assay showed no obvious difference between these two ACAT1 proteins respectively initiated from the two adjacent 5'-ORF-AUGs. This work showed that a stable upstream stem-loop structure could modulate the start codon selection during translation initiation of mRNAs that contain adjacent multi-5'-ORF-AUGs.

5' Untranslated Regions↗

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↗