Functional network motifs defined through integration of protein-protein and genetic interactions

Cells are enticingly complex systems. The identification of feedback regulation is critically important for understanding this complexity. Network motifs defined as small graphlets that occur more frequently than expected by chance have revolutionized our understanding of feedback circuits in cellular networks. However, with their definition solely based on statistical over-representation, network motifs often lack biological context, which limits their usefulness. Here, we define functional network motifs (FNMs) through the systematic integration of genetic interaction data that directly informs on functional relationships between genes and encoded proteins. Occurring two orders of magnitude less frequently than conventional network motifs, we found FNMs significantly enriched in genes known to be functionally related. Moreover, our comprehensive analyses of FNMs in yeast showed that they are powerful at capturing both known and putative novel regulatory interactions, thus suggesting a promising strategy towards the systematic identification of feedback regulation in biological networks. Many FNMs appeared as excellent candidates for the prioritization of follow-up biochemical characterization, which is a recurring bottleneck in the targeting of complex diseases. More generally, our work highlights a fruitful avenue for integrating and harnessing genomic network data.

TDP-43 proteinopathy alters the ribosome association of multiple mRNAs including the glypican Dally-like protein (Dlp)/GPC6

Amyotrophic lateral sclerosis (ALS) is a genetically heterogeneous neurodegenerative disease in which 97% of patients exhibit cytoplasmic aggregates containing the RNA binding protein TDP-43. Using tagged ribosome affinity purifications in Drosophila models of TDP-43 proteinopathy, we identified TDP-43 dependent translational alterations in motor neurons impacting the spliceosome, pentose phosphate and oxidative phosphorylation pathways. A subset of the mRNAs with altered ribosome association are also enriched in TDP-43 complexes suggesting that they may be direct targets. Among these, dlp mRNA, which encodes the glypican Dally like protein (Dlp)/GPC6, a wingless (Wg/Wnt) signaling regulator is insolubilized both in flies and patient tissues with TDP-43 pathology. While Dlp/GPC6 forms puncta in the Drosophila neuropil and ALS spinal cords, it is reduced at the neuromuscular synapse in flies suggesting compartment specific effects of TDP-43 proteinopathy. These findings together with genetic interaction data show that Dlp/GPC6 is a novel, physiologically relevant target of TDP-43 proteinopathy.

TDP-43 proteinopathy alters the ribosome association of multiple mRNAs including the glypican Dally-like protein (Dlp)/GPC6

Amyotrophic lateral sclerosis (ALS) is a genetically heterogeneous neurodegenerative disease in which 97% of patients exhibit cytoplasmic aggregates containing the RNA binding protein TDP-43. Using tagged ribosome affinity purifications in Drosophila models of TDP-43 proteinopathy, we identified TDP-43 dependent translational alterations in motor neurons impacting the spliceosome, pentose phosphate and oxidative phosphorylation pathways. A subset of the mRNAs with altered ribosome association are also enriched in TDP-43 complexes suggesting that they may be direct targets. Among these, dlp mRNA, which encodes the glypican Dally like protein (Dlp)/GPC6, a wingless (Wg/Wnt) signaling regulator is insolubilized both in flies and patient tissues with TDP-43 pathology. While Dlp/GPC6 forms puncta in the Drosophila neuropil and ALS spinal cords, it is reduced at the neuromuscular synapse in flies suggesting compartment specific effects of TDP-43 proteinopathy. These findings together with genetic interaction data show that Dlp/GPC6 is a novel, physiologically relevant target of TDP-43 proteinopathy.

NetR and AttR, Two New Bioinformatic Tools to Integrate Diverse Datasets into Cytoscape Network and Attribute Files

High-throughput technologies have allowed researchers to obtain genome-wide data from a wide array of experimental model systems. Unfortunately, however, new data generation tends to significantly outpace data re-utilization, and most high throughput datasets are only rarely used in subsequent studies or to generate new hypotheses to be tested experimentally. The reasons behind such data underutilization include a widespread lack of programming expertise among experimentalist biologists to carry out the necessary file reformatting that is often necessary to integrate published data from disparate sources. We have developed two programs (NetR and AttR), which allow experimental biologists with little to no programming background to integrate publicly available datasets into files that can be later visualized with Cytoscape to display hypothetical networks that result from combining individual datasets, as well as a series of published attributes related to the genes or proteins in the network. NetR also allows users to import protein and genetic interaction data from InterMine, which can further enrich a network model based on curated information. We expect that NetR/AttR will allow experimental biologists to mine a largely unexploited wealth of data in their fields and facilitate their integration into hypothetical models to be tested experimentally.

Efficient strategies for screening large-scale genetic interaction networks

Large-scale genetic interaction screening is a powerful approach for unbiased characterization of gene function and understanding systems-level cellular organization. While genome-wide screens are desirable as they provide the most comprehensive interaction profiles, they are resource and time-intensive and sometimes infeasible, depending on the species and experimental platform. For these scenarios, optimal methods for more efficient screening while still producing the maximal amount of information from the resulting profiles are of interest.To address this problem, we developed an optimal algorithm, called COMPRESS-GI, which selects a small but informative set of genes that captures most of the functional information contained within genome-wide genetic interaction profiles. The utility of this algorithm is demonstrated through an application of the approach to define a diagnostic mutant set for large-scale chemical genetic screens, where more than 13,000 compound screens were achieved through the increased throughput enabled by the approach. COMPRESS-GI can be broadly applied for directing genetic interaction screens in other contexts, including in species with little or no prior genetic-interaction data.

Improving prediction of compound function from chemical structure using chemical-genetic networks

ABSTRACTThe drug discovery process can be significantly improved through understanding how the structure of chemical compounds relates to their function. A common paradigm that has been used to filter and prioritize compounds is ligand-based virtual screening, where large libraries of compounds are queried for high structural similarity to a target molecule, with the assumption that structural similarity is predictive of similar biological activity. Although the chemical informatics community has already proposed a wide range of structure descriptors and similarity coefficients, a major challenge has been the lack of systematic and unbiased benchmarks for biological activity that covers a broad range of targets to definitively assess the performance of the alternative approaches.We leveraged a large set of chemical-genetic interaction data from the yeast Saccharomyces cerevisiae that our labs have recently generated, covering more than 13,000 compounds from the RIKEN NPDepo and several NCI, NIH, and GlaxoSmithKline (GSK) compound collections. Supportive of the idea that chemical-genetic interaction data provide an unbiased proxy for biological functions, we found that many commonly used structural similarity measures were able to predict the compounds that exhibited similar chemical-genetic interaction profiles, although these measures did exhibit significant differences in performance. Using the chemical-genetic interaction profiles as a basis for our evaluation, we performed a systematic benchmarking of 10 different structure descriptors, each combined with 12 different similarity coefficients. We found that the All-Shortest Path (ASP) structure descriptor paired with the Braun-Blanquet similarity coefficient provided superior performance that was robust across several different compound collections.We further describe a machine learning approach that improves the ability of the ASP metric to capture biological activity. We used the ASP fingerprints as input for several supervised machine learning models and the chemical-genetic interaction profiles as the standard for learning. We found that the predictive power of the ASP fingerprints (as well as several other descriptors) could be substantially improved by using support vector machines. For example, on held-out data, we measured a 5-fold improvement in the recall of biologically similar compounds at a precision of 50% based upon the ASP fingerprints. Our results generally suggest that using high-dimensional chemical-genetic data as a basis for refining chemical structure descriptors can be a powerful approach to improving prediction of biological function from structure.