Dataset Information

Improving the specificity of high-throughput ortholog prediction.

ABSTRACT:

Background

Orthologs (genes that have diverged after a speciation event) tend to have similar function, and so their prediction has become an important component of comparative genomics and genome annotation. The gold standard phylogenetic analysis approach of comparing available organismal phylogeny to gene phylogeny is not easily automated for genome-wide analysis; therefore, ortholog prediction for large genome-scale datasets is typically performed using a reciprocal-best-BLAST-hits (RBH) approach. One problem with RBH is that it will incorrectly predict a paralog as an ortholog when incomplete genome sequences or gene loss is involved. In addition, there is an increasing interest in identifying orthologs most likely to have retained similar function.

Results

To address these issues, we present here a high-throughput computational method named Ortholuge that further evaluates previously predicted orthologs (including those predicted using an RBH-based approach) - identifying which orthologs most closely reflect species divergence and may more likely have similar function. Ortholuge analyzes phylogenetic distance ratios involving two comparison species and an outgroup species, noting cases where relative gene divergence is atypical. It also identifies some cases of gene duplication after species divergence. Through simulations of incomplete genome data/gene loss, we show that the vast majority of genes falsely predicted as orthologs by an RBH-based method can be identified. Ortholuge was then used to estimate the number of false-positives (predominantly paralogs) in selected RBH-predicted ortholog datasets, identifying approximately 10% paralogs in a eukaryotic data set (mouse-rat comparison) and 5% in a bacterial data set (Pseudomonas putida - Pseudomonas syringae species comparison). Higher quality (more precise) datasets of orthologs, which we term "ssd-orthologs" (supporting-species-divergence-orthologs), were also constructed. These datasets, as well as Ortholuge software that may be used to characterize other species' datasets, are available at http://www.pathogenomics.ca/ortholuge/ (software under GNU General Public License).

Conclusion

The Ortholuge method reported here appears to significantly improve the specificity (precision) of high-throughput ortholog prediction for both bacterial and eukaryotic species. This method, and its associated software, will aid those performing various comparative genomics-based analyses, such as the prediction of conserved regulatory elements upstream of orthologous genes.

SUBMITTER: Fulton DL

PROVIDER: S-EPMC1524997 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundOne goal of structural biology is to understand how a protein's 3-dimensional conformation determines its capacity to interact with potential ligands. In the case of small chemical ligands, deconstructing a static protein-ligand complex into its constituent atom-atom interactions is typically sufficient to rapidly predict ligand affinity with high accuracy (>70% correlation between predicted and experimentally-determined affinity), a fact that is exploited to support structure-based drug design. We recently found that protein-DNA/RNA affinity can also be predicted with high accuracy using extensions of existing techniques, but protein-protein affinity could not be predicted with >60% correlation, even when the protein-protein complex was available.MethodsX-ray and NMR structures of protein-protein complexes, their associated binding affinities and experimental conditions were obtained from different binding affinity and structural databases. Statistical models were implemented using a generalized linear model framework, including the experimental conditions as new model features. We evaluated the potential for new features to improve affinity prediction models by calculating the Pearson correlation between predicted and experimental binding affinities on the training and test data after model fitting and after cross-validation. Differences in accuracy were assessed using two-sample t test and nonparametric Mann-Whitney U test.ResultsHere we evaluate a range of potential factors that may interfere with accurate protein-protein affinity prediction. We find that X-ray crystal resolution has the strongest single effect on protein-protein affinity prediction. Limiting our analyses to only high-resolution complexes (≤2.5 Å) increased the correlation between predicted and experimental affinity from 54 to 68% (p = 4.32x10-3). In addition, incorporating information on the experimental conditions under which affinities were measured (pH, temperature and binding assay) had significant effects on prediction accuracy. We also highlight a number of potential errors in large structure-affinity databases, which could affect both model training and accuracy assessment.ConclusionsThe results suggest that the accuracy of statistical models for protein-protein affinity prediction may be limited by the information present in databases used to train new models. Improving our capacity to integrate large-scale structural and functional information may be required to substantively advance our understanding of the general principles by which a protein's structure determines its function.

Project description:In this work, we used our recently developed method Spec-seq to characterize the binding specificity of Glucocorticoid receptor in vitro. This GR Spec-seq experiment has been run twice separately (Dec 2014 and Mar 2015) with different library compositions. The basic workflow is the same as our previous work for lac repressor published in Genetics 198.3 (2014): 1329-1343. Recombinant human GR protein was used to facilitate in vitro DNA-binding and separation experiments. Bound and Unbound DNA fragments were separated in EMSA gels, purified, barcoded for further Illumina sequencing. An initial experiment was performed using the putative consensus sequence: AGAACA GGG TGTTCT; randomized library 1: AGAACN NSN NGTTCT [diversity =512]; randomized library 2: AGAANN GGG NNNTCT [diversity = 1024]; randomized library 3: AGAACA GGG TGNNNN [diversity =256]; randomized library 4: AGAACA GGGC NNNTCT [diversity = 64]. The initial total library diversity was ~1853 with a library composition of 10% positive control sequence + randomized libraries 1-4 + 5% negative control sequence. A 5th library containing 2304 sequences based on DDNACW KKN KGTTCT, where D="not C", N="any base", W="A or T" and K="G or T" was subsequently prepared and analyzed using similar ratios of control sequences. Binding conditions for EMSA were 100ng FAM-labelled dsDNA+ 0/0.5/1/2/4uM GR DBD protein for each lane, 1X NEB buffer 4. GR DBD was prepared as previously described. The EMSA was performed using a 9% 33:1 acrylamide gel and TB buffer, and was run at 200V for 30mins @ 0 degrees C. The 2uM protein lane used for final sequencing. Bound/unbound fractions resulting from EMSA of these libraries and conditions were used to generate PWMs as described. The GR PWM that was generated through this analysis was used to define relative binding energies using the patser program (35), which can be accessed online at (http://stormo.wustl.edu/consensus/cgi-bin/Server/Interface/patser.cgi ). Derived binding affinities are proportional to the inverse of the natural log of the calculated energies.