Hanah Margalit

Background: Recently, there has been much interest in relating domain-domain interactions (DDIs) to protein-protein interactions (PPIs) and vice versa, in an attempt to understand the molecular basis of PPIs. Results: Here we map structurally derived DDIs onto the cellular PPI networks of different organisms and demonstrate that there is a catalog of domain pairs that is used to mediate various interactions in the cell. We show that these DDIs occur frequently in protein complexes and that homotypic interactions (of a domain with itself) are abundant. A comparison of the repertoires of DDIs in the networks of Escherichia coli, Saccharomyces cerevisiae, Caenorhabditis elegans, Drosophila melanogaster, and Homo sapiens shows that many DDIs are evolutionarily conserved. Conclusion: Our results indicate that different organisms use the same 'building blocks' for PPIs, suggesting that the functionality of many domain pairs in mediating protein interactions is maintained in evolution.

Databases of experimentally determined protein interactions provide information on binary interactions and on involvement in multiprotein complexes. These data are valuable for understanding the general properties of the interaction between proteins as well as for the development of prediction schemes for unknown interactions. Here we analyze experimentally determined protein interactions by measuring various sequence, genomic, transcriptomic, and proteomic attributes of each interacting pair in the yeast Saccharomyces cerevisiae . We find that dividing the data into two groups, one that includes binary interactions within protein complexes (stable) and another that includes binary interactions that are not within complexes (transient), enables better characterization of the interactions by the different attributes and improves the prediction of new interactions. This analysis revealed that most attributes were more indicative in the set of intracomplex interactions. Using this data set for training, we integrated the different attributes by logistic regression and developed a predictive scheme that distinguishes between interacting and noninteracting protein pairs. Analysis of the logistic-regression model showed that one of the strongest contributors to the discrimination between interacting and noninteracting pairs is the presence of distinct pairs of domain signatures that were suggested previously to characterize interacting proteins. The predictive algorithm succeeds in identifying both intracomplex and other interactions (possibly the more stable ones), and its correct identification rate is 2-fold higher than that of large-scale yeast two-hybrid experiments.

Background: Differences in the transcription regulation network are at the root of much of the phenotypic variation observed among organisms. These differences may be achieved either by changing the repertoire of regulators and/or their targets, or by rewiring the network. Following these changes and studying their logic is crucial for understanding the evolution of regulatory networks. Results: We use the well characterized transcription regulatory network of Escherichia coli K12 and follow the evolutionary changes in the repertoire of regulators and their targets across a large number of fully sequenced γ-proteobacteria. By focusing on close relatives of E. coli K12, we study the dynamics of the evolution of transcription regulation across a relatively short evolutionary timescale. We show significant differences in the evolution of repressors and activators. Repressors are only lost from a genome once their targets have themselves been lost, or once the network has significantly rewired. In contrast, activators are often lost even when their targets remain in the genome. As a result, E. coli K12 repressors that regulate many targets are rarely absent from organisms that are closely related to E. coli K12, while activators with a similar number of targets are often absent in these organisms. Conclusion: We demonstrate that the mode of regulation exerted by transcription factors has a strong effect on their evolution. Repressors co-evolve tightly with their target genes. In contrast, activators can be lost independently of their targets. In fact, loss of an activator can lead to efficient shutdown of an unnecessary pathway.

Current approaches for identification and detection of transcription factor binding sites rely on an extensive set of known target genes. Here we describe a novel structure-based approach applicable to transcription factors with no prior binding data. Our approach combines sequence data and structural information to infer context-specific amino acid–nucleotide recognition preferences. These are used to predict binding sites for novel transcription factors from the same structural family. We demonstrate our approach on the Cys₂His₂ Zinc Finger protein family, and show that the learned DNA-recognition preferences are compatible with experimental results. We use these preferences to perform a genome-wide scan for direct targets of Drosophila melanogaster Cys₂His₂ transcription factors. By analyzing the predicted targets along with gene annotation and expression data we infer the function and activity of these proteins. Cells respond to dynamic changes in their environment by invoking various cellular processes, coordinated by a complex regulatory program. A main component of this program is the regulation of transcription, which is mainly accomplished by transcription factors that bind the DNA in the vicinity of genes. To better understand transcriptional regulation, advanced computational approaches are needed for linking between transcription factors and their targets. The authors describe a novel approach by which the binding site of a given transcription factor can be characterized without previous experimental binding data. This approach involves learning a set of context-specific amino acid–nucleotide recognition preferences that, when combined with the sequence and structure of the protein, can predict its specific binding preferences. Applying this approach to the Cys₂His₂ Zinc Finger protein family demonstrated its genome-wide potential by automatically predicting the direct targets of 29 regulators in the genome of the fruit fly Drosophila melanogaster. At present, with the availability of many genome sequences, there are numerous proteins annotated as transcription factors based on their sequence alone. This approach offers a promising direction for revealing the targets of these factors and for understanding their roles in the cellular network.

We show that the predicted protein levels of functionally related proteins change in a coordinated fashion over many unicellular organisms. For each protein, we created a profile containing a protein abundance measure in each of a set of organisms. We show that for functionally related proteins these profiles tend to be correlated. Using the Codon Adaptation Index as a predictor of protein abundance in 48 unicellular organisms, we demonstrated this phenomenon for two types of functional relations: for proteins that physically interact and for proteins involved in consecutive steps within a metabolic pathway. Our results suggest that the protein abundance levels of functionally related proteins co-evolve.

MicroRNAs (miRNAs) are ∼22 nt-long non-coding RNA molecules, believed to play important roles in gene regulation. We present a comprehensive analysis of the conservation and clustering patterns of known miRNAs in human. We show that human miRNA gene clustering is significantly higher than expected at random. A total of 37% of the known human miRNA genes analyzed in this study appear in clusters of two or more with pairwise chromosomal distances of at most 3000 nt. Comparison of the miRNA sequences with their homologs in four other organisms reveals a typical conservation pattern, persistent throughout the clusters. Furthermore, we show enrichment in the typical conservation patterns and other miRNA-like properties in the vicinity of known miRNA genes, compared with random genomic regions. This may imply that additional, yet unknown, miRNAs reside in these regions, consistent with the current recognition that there are overlooked miRNAs. Indeed, by comparing our predictions with cloning results and with identified miRNA genes in other mammals, we corroborate the predictions of 18 additional human miRNA genes in the vicinity of the previously known ones. Our study raises the proportion of clustered human miRNAs that are <3000 nt apart to 42%. This suggests that the clustering of miRNA genes is higher than currently acknowledged, alluding to its evolutionary and functional implications.

Genes and proteins generate molecular circuitry that enables the cell to process information and respond to stimuli. A major challenge is to identify characteristic patterns in this network of interactions that may shed light on basic cellular mechanisms. Previous studies have analyzed aspects of this network, concentrating on either transcription–regulation or protein–protein interactions. Here we search for composite network motifs: characteristic network patterns consisting of both transcription–regulation and protein–protein interactions that recur significantly more often than in random networks. To this end we developed algorithms for detecting motifs in networks with two or more types of interactions and applied them to an integrated data set of protein–protein interactions and transcription regulation in Saccharomyces cerevisiae . We found a two-protein mixed-feedback loop motif, five types of three-protein motifs exhibiting coregulation and complex formation, and many motifs involving four proteins. Virtually all four-protein motifs consisted of combinations of smaller motifs. This study presents a basic framework for detecting the building blocks of networks with multiple types of interactions.

Protein expression in the cell is affected by various sequence-dependent features. Several such sequence-dependent features have been individually studied, yet they have not been compared quantitatively in terms of their relative influence on protein expression, and a hierarchy of these elements has not been determined. Here we present a quantitative analysis examining sequence-dependent features involved in prokaryotic translation, namely, the base-pairing potential between the mRNA Shine-Dalgarno sequence and the ribosomal RNA, codon bias, and the identity of the stop codon. We analyzed these features both at intra- and intergenomic levels using the Escherichia coli and Haemophilus influenzae genomes. Within each genome, we examined the relationship between each feature and protein expression levels determined by 2D-gel analyses. At the intergenomic level, comparative genomic principles were applied to study the relative preservation of the different sequence-dependent properties between orthologs. From these analyses, we determined that biased codon usage is the property that is most highly associated with protein expression and that is most conserved. The identity of the stop codon and the base-pairing potential of the mRNA Shine-Dalgarno sequence and the rRNA seem to have less of an effect on protein expression.

The post‐genomic era is marked by huge amounts of data generated by large‐scale functional genomic and proteomic experiments. A major challenge is to integrate the various types of genome‐scale information in order to reveal the intra‐ and inter‐ relationships between genes and proteins that constitute a living cell. Here we present a novel application of classical graph algorithms to integrate the cellular networks of protein–protein interactions and transcription regulation. We demonstrate how integration of these two networks enables the discovery of simple as well as complex regulatory circuits that involve both protein–protein and protein–DNA interactions. These circuits may serve for positive or negative feedback mechanisms. By applying our approach to data from the yeast Saccharomyces cerevisiae, we were able to identify known simple and complex regulatory circuits and to discover many putative circuits, whose biological relevance has been assessed using various types of experimental data. The newly identified relations provide new insight into the processes that take place in the cell, insight that could not be gained by analyzing each type of data independently. The computational scheme that we propose may be used to integrate additional functional genomic and proteomic data and to reveal other types of relations, in yeast as well as in higher organisms.