Sequence-based prediction of protein-protein interactions by means of codon usage
© Najafabadi and Salavati; licensee BioMed Central Ltd. 2008
Received: 6 February 2008
Accepted: 23 May 2008
Published: 23 May 2008
We introduce a novel approach to predict interaction of two proteins solely by analyzing their coding sequences. We found that similarity in codon usage is a strong predictor of protein-protein interactions and, for high specificity values, is as sensitive as the most powerful current prediction methods. Furthermore, combining codon usage with other predictors results in a 75% increase in sensitivity at a precision of 50%, compared to prediction without considering codon usage.
The need to transform the growing amount of biological information into knowledge has involved several disciplines that, by means of experimental and computational approaches, aim to decipher functional linkages and interactions between proteins [1, 2]. Current computational methods for predicting protein-protein interactions demand data that, compared to the huge amount of available genomic sequences, are scarce. Only in a few organisms have features such as essentiality, biological function and mRNA co-expression of genes been partially determined. Also, a combination of different homology-based predictors, including phylogenetic profiles , Rosetta stone  and interolog mapping , has provided incomplete information about interactions of only one-third of all Saccharomyces cerevisiae proteins. Hence, a method to identify protein-protein interactions solely on the basis of gene sequences would significantly expand the ability to predict interaction networks.
A few studies have been performed on the prediction of protein-protein interactions based only on amino acid sequence information [6–8]. However, the highest specificity reported in these studies is 86%. Considering the number of possible protein pairs in a genome consisting of no more than 6,000 protein-coding genes, this level of specificity results in the unacceptable number of 2.5 × 106 false positives. These studies consider protein sequences, and ignore the plethora of information that exists in their coding sequences. The still-unsatisfied demand for reliable sequence-based prediction of protein-protein interactions encourages exploration of relevant sequence features in the genome instead of the proteome.
It has been widely acknowledged that codon usage is correlated with expression level . In addition, it has been shown that codon usage is structured along the genome , with near neighbor genes having similar codon compositions. Some function-specific codon preferences have also been hypothesized based on selective charging of tRNA isoacceptors  and have been confirmed experimentally . Based on these premises and considering that similarity in mRNA expression pattern and biological function, along with physical gene proximity, are powerful predictors of protein-protein interactions , codon usage can be considered as a potential candidate for analysis. The coevolution of codon usage of functionally linked genes has been explicitly reported before [14, 15]. These studies suggest that the codon adaptation index (CAI)  of functionally related proteins changes in a coordinated fashion over different unicellular organisms. However, identification of this coordination between two genes needs the presence of orthologues in several organisms; hence, many species-specific genes, which are usually the hot spots of attraction for biologists, are excluded. Also, there are genes with very low variation in the CAI over different organisms , for which this kind of analysis is unreliable.
In this paper, we show that codon usage of functionally and/or physically linked proteins in an organism contain enough information to enable us to detect these proteins, even in the absence of homologues in other organisms. Furthermore, we show that our method is several times more sensitive than tracking the coordinated changes of codon usage over different organisms, and in fact is one of the best methods for identification of protein-protein interactions.
Results and discussion
Gold standard sets
No. of ORFs
No. of ORF pairs
Derived from MIPS  complex catalog. We excluded ribosomal proteins to avoid bias towards extreme codon usage similarity of their genes
Pairs of proteins that are not localized in the same cell compartment. We excluded ribosomal proteins
Protein pairs within the same KEGG  pathway
Protein pairs with KEGG information, excluding pairs in the gold standard positive set
Pull-down assay using a His-tagged ORF library
We compiled a set of protein pairs that are not in the gold standard positive set, given that at least one protein from each pair is copurified with an associate protein by Arifuzzaman et al. 
d ij (c) = |f i (c) - f j (c)|
where f i (c) and f j (c) are relative frequencies of codon c in ORF i and ORF j, respectively (Σ k f i (c k ) = 1 and Σ k f j (c k ) = 1; k = 1,2,..64 indicates all 64 codons). Therefore, d ij demonstrates the distance of two ORFs in terms of usage of codon c. We found that for almost all codons, distribution of d differed between positive and negative gold standard sets (Additional data file 1). Generally, distribution of d shifts to smaller values for ORFs within the gold standard positive set, indicating that interacting ORFs are more similar in codon usage profile than non-interacting ORFs. However, this shift is marginal for each codon individually, which means that single codons are weak predictors of protein-protein interactions.
We divided the distribution of d for each codon into 50 intervals, for each of which we calculated the likelihood ratio, that is, the fraction of positive gold standards occurring in that interval divided by the fraction of negatives occurring in that interval. Since the mutual information of d for each pair of codons was negligible, we combined these likelihood ratios using a naïve Bayes approach (see Additional data files 2 and 3 for a graphical representation). Although obviously not all features were independent from each other (with statistical tests suggesting 10 to 16 independent components; see Additional data file 4), we found that a naïve Bayesian network is more effective than a Bayesian network in which each variable node has one other parent node, perhaps because the increase of the parameters in the latter case causes partial overfitting of the network. Using a tenfold cross-validation method, we evaluated the performance of this naïve Bayesian network in predicting protein-protein interactions. To do so, we divided the gold-standard set into ten random segments; each time we used nine segments as the training set and calculated the combined likelihood ratios for each ORF pair in the remaining segment. We designate the method 'PIC' (for probabilistic-interactome using codon usage).
Although PIC considers the relative frequencies of codons in ORF pairs, it reflects not only synonymous codon usage, but also amino acid frequencies and ORF lengths. ORF length is reflected in PIC since stop codons are not omitted, and each ORF has only one stop codon. Therefore, the relative frequency of a stop codon in long ORFs is smaller than in short ORFs. We created three other probabilistic interaction networks of S. cerevisiae using RSCU , relative frequencies of amino acids, and ORF length in order to examine the effect of each factor. We named these probabilistic networks PI-RSCU, PI-A and PI-L, respectively. RSCU is a measure of synonymous codon usage that is independent of amino acid composition (see reference  for the definition of RSCU and to compare it with the relative frequency of codon. RSCU as well as many other measures of synonymous codon usage are dependent on gene length, and result in biased values when the corresponding coding sequences are short . In the worst case, when an amino acid is absent from a gene, it is impossible to calculate the RSCU for its corresponding codons. In the latter case, we treated the RSCU values of these codons as missing data, which can be easily handled by naïve Bayesian networks. In comparable sensitivities, the descending order of accuracy was PIC > PI-RSCU > PI-A > PI-L (Additional data file 5). This suggests a synergistic effect of each of these factors on the strength of PIC, with synonymous codon usage being the most important one. It should be mentioned that the length of the protein (PI-L) has a very marginal ability to distinguish interacting from non-interacting pairs, and even this observed marginal prediction may be due to the bias of the gold standard positive set towards a certain range of protein lengths, as the length of a protein affects many experimental procedures, such as successful cloning, and so on.
PIC can easily be combined with other probabilistic approaches, such as PIP (PI-predicted) and PIT (PI-total)  (see Materials and methods for combining two probabilistic interactomes). PIP is a probabilistic predicted network of S. cerevisiae in which four datasets of genomic features are integrated: two datasets of biological functions, a dataset of mRNA expression correlation and a dataset of essentiality . Jansen et al.  showed that, at comparable levels of sensitivity, PIP is even more accurate than PIE (PI-experimental), a probabilistic network constructed by integration of four experimental datasets of the yeast interactome. They also combined PIP and PIE into PIT as one of the most comprehensive probabilistic networks of known and putative protein complexes in yeast. We integrated the results of yeast PIC and PIP to see how their combination improves our power in de novo prediction of interactions.
Finally, we combined PIT  and PIC to generate 'PICT', which we propose as one of the most reliable probabilistic interactomes of S. cerevisiae (see Additional data file 11 for precision-recall curves of PIT and PICT. PICT, accompanied by PIC for the whole genome of S. cerevisiae, is available online ). At a likelihood cutoff of 2 × 103, PICT has the same specificity as PIT-Lcut600, while, after excluding promiscuous nodes (that is, nodes each of which has ≥100 edges), it includes 1,306 more ORFs compared to PIT. Analysis of PICT-Lcut2000 reveals many interesting interactions not present in PIT-Lcut600. Some examples are represented below. We specifically consider complexes that were also examined by Jansen et al.  in order to provide a more detailed comparison between PIT and PICT. Note that the following interactions should be considered as complex co-memberships rather than direct physical interactions, since all the components of PICT are trained on protein complexes and not one-to-one physical interactions of proteins. However, a direct physical interaction is also possible based on the closeness of proteins within the same complex.
Another example is the case of translation initiation/elongation factors. PIT-Lcut600 fails to predict an interaction involving elongation factor 2 (EF-2). It also predicts only two interactions for EF-1α, with EF-1β and EF-1γ. Although PIT-Lcut300 suggests some more interactions for these proteins, a higher rate of false positives in PIT-Lcut300 renders them unreliable. PICT-Lcut2000 predicts several interactions involving different elongation factors as well as initiation factors 4A and 5A, many of which have been recently confirmed by tandem-affinity purification experiments [22, 26–29]]. Figure 4 shows a subgraph of PICT-Lcut2000 representing interactions among translation initiation/elongation factors that are not present in PIT-Lcut600. A recent study  has shown that Poly(A)-binding protein Pab1 interacts with EF-1α. Based on PICT-Lcut2000, we anticipate that Pab1 interacts with EF-2 and EF-1γ as well. Also, we found an interesting interaction between the ribosome-associated molecular chaperone Ssb1 and eIF4A. Interaction of Ssb1 and eIF4G has already been shown by tandem-affinity purification . Based on the close interaction of eIF4A and eIF4G, interaction of Ssb1 and eIF4A is reasonable.
RNase P complex represents another interesting example of PICT predictions. PICT-Lcut2000 predicts six new interactions between RNase P complex and other proteins in yeast, neither of which exists in PIT-Lcut600 or has been reported previously. Four interactions are with uncharacterized ORFs, YKL096C-B, YDL159W-A, YKL183C-A and Q0255. Q0255 is likely to code for a maturase-like protein. It has been hypothesized that mitochondrial maturases participate in splicing by stabilizing some secondary or tertiary structure needed for splicing . Their exact function, however, remains uncharacterized . An interaction between RNase P complex and Q0255 implies the plausibility that this protein could contribute to maturation of ribosomal RNA and tRNA in mitochondria. According to PICT-Lcut2000, HUB1 (Histone mono-ubiquitination 1) is another interacting partner of RNase P complex. Previous data have shown that HUB1 is a functional homolog of the human and yeast BRE1 proteins, and suggest that it mediates gene activation and cell cycle regulation through chromatin modifications . In addition, chromatin remodeling in Arabidopsis thaliana seed dormancy is proposed to be mediated by H2B mono-ubiquitination through HUB1 and HUB2 . In agreement with this, the recently reported binding of human RNase P to chromatin of non-coding RNA genes and regulation of pol III transcription  could be mediated through a HUB1-RNase P interaction. Another prediction of PICT-Lcut2000, interaction of RNase P with CKB1, also corroborates this observation. CKB1 is a regulatory subunit of casein kinase 2, whose many substrates include transcription factors and all RNA polymerases. Again, this is consistent with the recent proposed role for RNase P in pol III transcription [33, 34].
We notice that PICT has the potential of providing new information about proteins that lack homology. For example, YAR068W is a fungal-specific gene, for which PIT has no interaction. This is while PICT predicts an interaction between this protein and a protein of the large subunit of mitochondrial ribosome (refer to PICT-Lcut2000 in Additional data file 13).
PIC uses a naïve Bayesian network to combine the information provided by the frequencies of all codons in order to predict protein-protein interactions. Given a set of independent features, naïve Bayesian networks can combine them in a way that minimizes the loss of information that usually occurs by the aggregation of several features. Depending on the training set that has been used, PIC can predict both complex membership (as in the Munich Information Center for Protein Sequences (MIPS) database or TAP-tagging experiments) and functional linkages between proteins (as in the KEGG pathway database). Although we did not test the power of PIC for prediction of direct physical interactions between proteins, it is possible that it can be used for that purpose as well, since complex membership, functional linkage and direct physical interactions are all properties that are highly inter-correlated. We anticipate that integrating PIC with the current knowledge of protein interactions in different organisms will significantly increase the reliability and coverage of probabilistic interactomes. In the case of S. cerevisiae, the results of PIC as well as its combination with PIT , referred to in this article as PICT, are provided online . This study not only describes a novel method for de novo prediction of protein-protein interactions, but also suggests the plausibility of previously unseen evolutionary forces acting on codon compositions of genes within a genome. A few studies have taken into account the effect of protein-protein interactions on codon usage; however, these studies generally consider the unique features of codon composition of an ORF in regions that code the interacting face of the protein compared to the rest of the ORF , not the direct relationship between codon usages of two interacting proteins. Characterization of evolutionary mechanisms shaping these relationships may lead to development of even more powerful methods for sequence-based prediction of interaction networks.
Materials and methods
Analysis of genomic features
We used d ij (ζ k ) = |ζ k i - ζ k j | to measure the distance of two genes i and j regarding feature ζ k . In the case of PIC, ζ k = f(c k ), where f(c k ) is the normalized frequency of usage of codon c k , so that Σ k f (c k ) = 1 (1 ≤ k ≤ 64). For PI-RSCU, ζ k = RSCU(c k ) (see ). For PI-A, ζ k = f(a k ), where f(a k ) is the normalized frequency of amino acid a k (1 ≤ k ≤ 20). For PI-L, ζ = L, where L represents the ORF length. To combine a set of features, a naïve Bayesian network  is employed. Naïve Bayesian networks are most effective when they are used to combine independent features. We assessed independency of d ij for two features r and s by means of mutual information , where I [d ij (ζ r );d ij (ζ s )] < 0.01 was assumed not to influence the performance of the naïve Bayesian network. To combine two probabilistic networks, we multiplied the likelihoods each network assigned to each interaction.
Coevolution of CAI
We performed the same analysis as described by Fraser et al. , using the genome sequences of S. cerevisiae, Saccharomyces paradoxus, Saccharomyces mikatae, and Saccharomyces bayanus . We used species-specific adaptation index to determine the CAI values by using the codon frequencies of the 20 most highly expressed genes. We assumed that the 20 most highly expressed genes in the four species are the same; hence, we used a previous report on mRNA expression in S. cerevisiae  to identify them. Addition of E. coli in the analysis did not improve the results. We did not add more genomes because we would lose a portion of our gold standard sets, especially the negative gold standard set, due to the lack of homology for all genes among all genomes, resulting in non-comparable sensitivity/specificity values.
Additional data files
The following additional data are available with the online version of this paper. Additional data file 1 is a figure showing the distribution of d for each codon in yeast. Additional data file 2 is a figure comparing the naïve Bayesian network and fully connected Bayesian network in the yeast gold standard positive set. Additional data file 3 is a figure comparing the naïve Bayesian network and fully connected Bayesian network in the yeast gold standard negative set. Additional data file 4 demonstrates the variance over different components resulting from principal component analysis of the interacting gene pairs in yeast. Additional data file 5 compares PIC, PI-RSCU, PI-A and PI-L in a figure. Additional data file 6 is a figure comparing PIP × PIC and the yeast gold standard positive set. Additional data file 7 illustrates per-complex comparison of PIP and PIP × PIC in a figure. Additional data file 8 is a figure showing the MIPS functional category enrichment for the yeast genome, PIP-Lcut600 and PIC-Lcut600. Additional data file 9 is a figure representing the distribution of mRNA expression levels in interactions predicted by PIP-Lcut600 and PIC-Lcut600 for S. cerevisiae. Additional data file 10 shows the precision-recall curves for PIC, interolog mapping (INT), phylogenetic profiles (PGP), Rosetta stone (ROS), CAI coevolution (co-CAI) and CAI. Additional data file 11 includes precision-recall curves for PIC, PIT and PICT. Additional data file 12 is a compressed file containing PIC-Lcut600 for S. cerevisiae. Additional data file 13 is a compressed file containing PICT-Lcut2000 for S. cerevisiae. Additional data file 14 is a compressed file containing the results of performance of PIC on the P. falciparum gold standard set.
codon adaptation index
high mobility group
Kyoto Encyclopedia of Genes and Genomes
Munich Information Center for Protein Sequences
open reading frame
PI using amino acid frequencies
probabilistic-interactome using codon usage
combination of PIC and PIT
PI using sequence length
PI using RSCU
We thank TG Geary and K Hassani for reading the manuscript and for their critical comments. We would also like to thank two anonymous referees for their constructive suggestions. Research at the Institute of Parasitology is supported by the Centre for Host-Parasite Interactions and Le Fonds quebecois de la recherche sur la nature et les technologies (FQRNT), Quebec. HSN is supported by the Max Stern Fellowship from McGill University and a fellowship from McGill Centre for Bioinformatics.
- Eisenberg D, Marcotte EM, Xenarios I, Yeates TO: Protein function in the post-genomic era. Nature. 2000, 405: 823-826. 10.1038/35015694.PubMedView ArticleGoogle Scholar
- Sharan R, Ulitsky I, Shamir R: Network-based prediction of protein function. Molecular systems biology. 2007, 3: 88-10.1038/msb4100129.PubMedPubMed CentralView ArticleGoogle Scholar
- Pellegrini M, Marcotte EM, Thompson MJ, Eisenberg D, Yeates TO: Assigning protein functions by comparative genome analysis: protein phylogenetic profiles. Proceedings of the National Academy of Sciences of the United States of America. 1999, 96: 4285-4288. 10.1073/pnas.96.8.4285.PubMedPubMed CentralView ArticleGoogle Scholar
- Marcotte EM, Pellegrini M, Ng HL, Rice DW, Yeates TO, Eisenberg D: Detecting protein function and protein-protein interactions from genome sequences. Science. 1999, 285: 751-753. 10.1126/science.285.5428.751.PubMedView ArticleGoogle Scholar
- Yu H, Luscombe NM, Lu HX, Zhu X, Xia Y, Han JD, Bertin N, Chung S, Vidal M, Gerstein M: Annotation transfer between genomes: protein-protein interologs and protein-DNA regulogs. Genome research. 2004, 14: 1107-1118. 10.1101/gr.1774904.PubMedPubMed CentralView ArticleGoogle Scholar
- Shen J, Zhang J, Luo X, Zhu W, Yu K, Chen K, Li Y, Jiang H: Predicting protein-protein interactions based only on sequences information. Proceedings of the National Academy of Sciences of the United States of America. 2007, 104: 4337-4341. 10.1073/pnas.0607879104.PubMedPubMed CentralView ArticleGoogle Scholar
- Bock JR, Gough DA: Predicting protein - protein interactions from primary structure. Bioinformatics (Oxford, England). 2001, 17: 455-460. 10.1093/bioinformatics/17.5.455.View ArticleGoogle Scholar
- Nanni L, Lumini A: An ensemble of K-local hyperplanes for predicting protein-protein interactions. Bioinformatics (Oxford, England). 2006, 22: 1207-1210. 10.1093/bioinformatics/btl055.View ArticleGoogle Scholar
- Jansen R, Bussemaker HJ, Gerstein M: Revisiting the codon adaptation index from a whole-genome perspective: analyzing the relationship between gene expression and codon occurrence in yeast using a variety of models. Nucleic acids research. 2003, 31: 2242-2251. 10.1093/nar/gkg306.PubMedPubMed CentralView ArticleGoogle Scholar
- Daubin V, Perriere G: G+C3 structuring along the genome: a common feature in prokaryotes. Molecular biology and evolution. 2003, 20: 471-483. 10.1093/molbev/msg022.PubMedView ArticleGoogle Scholar
- Elf J, Nilsson D, Tenson T, Ehrenberg M: Selective charging of tRNA isoacceptors explains patterns of codon usage. Science. 2003, 300: 1718-1722. 10.1126/science.1083811.PubMedView ArticleGoogle Scholar
- Dittmar KA, Sorensen MA, Elf J, Ehrenberg M, Pan T: Selective charging of tRNA isoacceptors induced by amino-acid starvation. EMBO reports. 2005, 6: 151-157. 10.1038/sj.embor.7400341.PubMedPubMed CentralView ArticleGoogle Scholar
- Lu LJ, Xia Y, Paccanaro A, Yu H, Gerstein M: Assessing the limits of genomic data integration for predicting protein networks. Genome research. 2005, 15: 945-953. 10.1101/gr.3610305.PubMedPubMed CentralView ArticleGoogle Scholar
- Fraser HB, Hirsh AE, Wall DP, Eisen MB: Coevolution of gene expression among interacting proteins. Proceedings of the National Academy of Sciences of the United States of America. 2004, 101: 9033-9038. 10.1073/pnas.0402591101.PubMedPubMed CentralView ArticleGoogle Scholar
- Lithwick G, Margalit H: Relative predicted protein levels of functionally associated proteins are conserved across organisms. Nucleic acids research. 2005, 33: 1051-1057. 10.1093/nar/gki261.PubMedPubMed CentralView ArticleGoogle Scholar
- Sharp PM, Li WH: The codon Adaptation Index - a measure of directional synonymous codon usage bias, and its potential applications. Nucleic acids research. 1987, 15: 1281-1295. 10.1093/nar/15.3.1281.PubMedPubMed CentralView ArticleGoogle Scholar
- Musto H, Romero H, Zavala A, Jabbari K, Bernardi G: Synonymous codon choices in the extremely GC-poor genome of Plasmodium falciparum : compositional constraints and translational selection. Journal of molecular evolution. 1999, 49: 27-35. 10.1007/PL00006531.PubMedView ArticleGoogle Scholar
- Enright AJ, Iliopoulos I, Kyrpides NC, Ouzounis CA: Protein interaction maps for complete genomes based on gene fusion events. Nature. 1999, 402: 86-90. 10.1038/47056.PubMedView ArticleGoogle Scholar
- Kanehisa M, Goto S, Hattori M, Aoki-Kinoshita KF, Itoh M, Kawashima S, Katayama T, Araki M, Hirakawa M: From genomics to chemical genomics: new developments in KEGG. Nucleic acids research. 2006, 34: D354-357. 10.1093/nar/gkj102.PubMedPubMed CentralView ArticleGoogle Scholar
- Sharp PM, Tuohy TM, Mosurski KR: Codon usage in yeast: cluster analysis clearly differentiates highly and lowly expressed genes. Nucleic acids research. 1986, 14: 5125-5143. 10.1093/nar/14.13.5125.PubMedPubMed CentralView ArticleGoogle Scholar
- Fuglsang A: Estimating the "effective number of codons": the Wright way of determining codon homozygosity leads to superior estimates. Genetics. 2006, 172: 1301-1307. 10.1534/genetics.105.049643.PubMedPubMed CentralView ArticleGoogle Scholar
- Jansen R, Yu H, Greenbaum D, Kluger Y, Krogan NJ, Chung S, Emili A, Snyder M, Greenblatt JF, Gerstein M: A Bayesian networks approach for predicting protein-protein interactions from genomic data. Science. 2003, 302: 449-453. 10.1126/science.1087361.PubMedView ArticleGoogle Scholar
- Genome wide analysis of yeast interactome with PIC and PICT and analysis of E. coli gold standard set with PIC. [http://webpages.mcgill.ca/staff/Group2/rsalav/web/]
- Krogan NJ, Cagney G, Yu H, Zhong G, Guo X, Ignatchenko A, Li J, Pu S, Datta N, Tikuisis AP, Punna T, Peregrin-Alvarez JM, Shales M, Zhang X, Davey M, Robinson MD, Paccanaro A, Bray JE, Sheung A, Beattie B, Richards DP, Canadien V, Lalev A, Mena F, Wong P, Starostine A, Canete MM, Vlasblom J, Wu S, Orsi C, et al: Global landscape of protein complexes in the yeast Saccharomyces cerevisiae. Nature. 2006, 440: 637-643. 10.1038/nature04670.PubMedView ArticleGoogle Scholar
- Huh WK, Falvo JV, Gerke LC, Carroll AS, Howson RW, Weissman JS, O'Shea EK: Global analysis of protein localization in budding yeast. Nature. 2003, 425: 686-691. 10.1038/nature02026.PubMedView ArticleGoogle Scholar
- Collins SR, Kemmeren P, Zhao XC, Greenblatt JF, Spencer F, Holstege FC, Weissman JS, Krogan NJ: Toward a comprehensive atlas of the physical interactome of Saccharomyces cerevisiae. Mol Cell Proteomics. 2007, 6: 439-450.PubMedView ArticleGoogle Scholar
- Gavin AC, Aloy P, Grandi P, Krause R, Boesche M, Marzioch M, Rau C, Jensen LJ, Bastuck S, Dumpelfeld B, Edelmann A, Heurtier MA, Hoffman V, Hoefert C, Klein K, Hudak M, Michon AM, Schelder M, Schirle M, Remor M, Rudi T, Hooper S, Bauer A, Bouwmeester T, Casari G, Drewes G, Neubauer G, Rick JM, Kuster B, Bork P, et al: Proteome survey reveals modularity of the yeast cell machinery. Nature. 2006, 440: 631-636. 10.1038/nature04532.PubMedView ArticleGoogle Scholar
- Jao DL, Chen KY: Tandem affinity purification revealed the hypusine-dependent binding of eukaryotic initiation factor 5A to the translating 80S ribosomal complex. Journal of cellular biochemistry. 2006, 97: 583-598. 10.1002/jcb.20658.PubMedView ArticleGoogle Scholar
- Anand M, Chakraburtty K, Marton MJ, Hinnebusch AG, Kinzy TG: Functional interactions between yeast translation eukaryotic elongation factor (eEF) 1A and eEF3. The Journal of biological chemistry. 2003, 278: 6985-6991. 10.1074/jbc.M209224200.PubMedView ArticleGoogle Scholar
- Hebbar SK, Belcher SM, Perlman PS: A maturase-encoding group IIA intron of yeast mitochondria self-splices in vitro. Nucleic acids research. 1992, 20: 1747-1754. 10.1093/nar/20.7.1747.PubMedPubMed CentralView ArticleGoogle Scholar
- Belfort M: Two for the price of one: a bifunctional intron-encoded DNA endonuclease-RNA maturase. Genes & development. 2003, 17: 2860-2863. 10.1101/gad.1162503.View ArticleGoogle Scholar
- Fleury D, Himanen K, Cnops G, Nelissen H, Boccardi TM, Maere S, Beemster GT, Neyt P, Anami S, Robles P, Micol JL, Inze D, Van Lijsebettens M: The Arabidopsis thaliana homolog of yeast BRE1 has a function in cell cycle regulation during early leaf and root growth. The Plant cell. 2007, 19: 417-432. 10.1105/tpc.106.041319.PubMedPubMed CentralView ArticleGoogle Scholar
- Jarrous N, Reiner R: Human RNase P: a tRNA-processing enzyme and transcription factor. Nucleic acids research. 2007, 35: 3519-3524. 10.1093/nar/gkm071.PubMedPubMed CentralView ArticleGoogle Scholar
- Reiner R, Ben-Asouli Y, Krilovetzky I, Jarrous N: A role for the catalytic ribonucleoprotein RNase P in RNA polymerase III transcription. Genes & development. 2006, 20: 1621-1635. 10.1101/gad.386706.View ArticleGoogle Scholar
- Alvarez-Valin F, Tort JF, Bernardi G: Nonrandom spatial distribution of synonymous substitutions in the GP63 gene from Leishmania. Genetics. 2000, 155: 1683-1692.PubMedPubMed CentralGoogle Scholar
- Saccharomyces cerevisiae coding sequences. [ftp://genome-ftp.stanford.edu/pub/yeast/sequence/genomic_sequence/orf_dna/orf_coding_all.fasta.gz]
- Escherichia coli coding sequences. [http://gib.genes.nig.ac.jp/pub/Ecol_K12_W3110/Ecol_K12_W3110_.ena.gz]
- Plasmodium falciparum coding sequences. [http://www.plasmodb.org/common/downloads/release-5.4/Pfalciparum/PfalciparumAnnotatedCDS_plasmoDB-5.4.fasta]
- Kellis M, Patterson N, Endrizzi M, Birren B, Lander ES: Sequencing and comparison of yeast species to identify genes and regulatory elements. Nature. 2003, 423: 241-254. 10.1038/nature01644.PubMedView ArticleGoogle Scholar
- Greenbaum D, Jansen R, Gerstein M: Analysis of mRNA expression and protein abundance data: an approach for the comparison of the enrichment of features in the cellular population of proteins and transcripts. Bioinformatics (Oxford, England). 2002, 18: 585-596. 10.1093/bioinformatics/18.4.585.View ArticleGoogle Scholar
- Saccharomyces cerevisiae Bayesian features. [http://networks.gersteinlab.org/BayesFeatures/]
- Mewes HW, Amid C, Arnold R, Frishman D, Guldener U, Mannhaupt G, Munsterkotter M, Pagel P, Strack N, Stumpflen V, Warfsmann J, Ruepp A: MIPS: analysis and annotation of proteins from whole genomes. Nucleic acids research. 2004, 32: D41-44. 10.1093/nar/gkh092.PubMedPubMed CentralView ArticleGoogle Scholar
- Date SV, Stoeckert CJ: Computational modeling of the Plasmodium falciparum interactome reveals protein function on a genome-wide scale. Genome research. 2006, 16: 542-549. 10.1101/gr.4573206.PubMedPubMed CentralView ArticleGoogle Scholar
- Arifuzzaman M, Maeda M, Itoh A, Nishikata K, Takita C, Saito R, Ara T, Nakahigashi K, Huang HC, Hirai A, Tsuzuki K, Nakamura S, Altaf-Ul-Amin M, Oshima T, Baba T, Yamamoto N, Kawamura T, Ioka-Nakamichi T, Kitagawa M, Tomita M, Kanaya S, Wada C, Mori H: Large-scale identification of protein-protein interaction of Escherichia coli K-12. Genome research. 2006, 16: 686-691. 10.1101/gr.4527806.PubMedPubMed CentralView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.