Siri of the cell: what biology could learn from the iPhone.

Modern genomics is very efficient at mapping genes and gene networks, but how to transform these maps into predictive models of the cell remains unclear. Recent progress in computer science, embodied by intelligent agents such as Siri, inspires an approach for moving from networks to multiscale models able to predict a range of cellular phenotypes and answer biological questions.

A proteome-scale map of the human interactome network.

Just as reference genome sequences revolutionized human genetics, reference maps of interactome networks will be critical to fully understand genotype-phenotype relationships. Here, we describe a systematic map of ?14,000 high-quality human binary protein-protein interactions. At equal quality, this map is ?30% larger than what is available from small-scale studies published in the literature in the last few decades. While currently available information is highly biased and only covers a relatively small portion of the proteome, our systematic map appears strikingly more homogeneous, revealing a "broader" human interactome network than currently appreciated. The map also uncovers significant interconnectivity between known and candidate cancer gene products, providing unbiased evidence for an expanded functional cancer landscape, while demonstrating how high-quality interactome models will help "connect the dots" of the genomic revolution.

Biological factors and statistical limitations prevent detection of most noncanonical proteins by mass spectrometry.

Ribosome profiling experiments indicate pervasive translation of short open reading frames (ORFs) outside of annotated protein-coding genes. However, shotgun mass spectrometry (MS) experiments typically detect only a small fraction of the predicted protein products of this noncanonical translation. The rarity of detection could indicate that most predicted noncanonical proteins are rapidly degraded and not present in the cell; alternatively, it could reflect technical limitations. Here, we leveraged recent advances in ribosome profiling and MS to investigate the factors limiting detection of noncanonical proteins in yeast. We show that the low detection rate of noncanonical ORF products can largely be explained by small size and low translation levels and does not indicate that they are unstable or biologically insignificant. In particular, proteins encoded by evolutionarily young genes, including those with well-characterized biological roles, are too short and too lowly expressed to be detected by shotgun MS at current detection sensitivities. Additionally, we find that decoy biases can give misleading estimates of noncanonical protein false discovery rates, potentially leading to false detections. After accounting for these issues, we found strong evidence for 4 noncanonical proteins in MS data, which were also supported by evolution and translation data. These results illustrate the power of MS to validate unannotated genes predicted by ribosome profiling, but also its substantial limitations in finding many biologically relevant lowly expressed proteins.

On the illusion of auxotrophy: met15Δ yeast cells can grow on inorganic sulfur, thanks to the previously uncharacterized homocysteine synthase Yll058w.

Organisms must either synthesize or assimilate essential organic compounds to survive. The homocysteine synthase Met15 has been considered essential for inorganic sulfur assimilation in yeast since its discovery in the 1970s. As a result, MET15 has served as a genetic marker for hundreds of experiments that play a foundational role in eukaryote genetics and systems biology. Nevertheless, we demonstrate here through structural and evolutionary modeling, in vitro kinetic assays, and genetic complementation, that an alternative homocysteine synthase encoded by the previously uncharacterized gene YLL058W enables cells lacking Met15 to assimilate enough inorganic sulfur for survival and proliferation. These cells however fail to grow in patches or liquid cultures unless provided with exogenous methionine or other organosulfurs. We show that this growth failure, which has historically justified the status of MET15 as a classic auxotrophic marker, is largely explained by toxic accumulation of the gas hydrogen sulfide because of a metabolic bottleneck. When patched or cultured with a hydrogen sulfide chelator, and when propagated as colony grids, cells without Met15 assimilate inorganic sulfur and grow, and cells with Met15 achieve even higher yields. Thus, Met15 is not essential for inorganic sulfur assimilation in yeast. Instead, MET15 is the first example of a yeast gene whose loss conditionally prevents growth in a manner that depends on local gas exchange. Our results have broad implications for investigations of sulfur metabolism, including studies of stress response, methionine restriction, and aging. More generally, our findings illustrate how unappreciated experimental variables can obfuscate biological discovery.

Elastic network modeling of cellular networks unveils sensor and effector genes that control information flow.

The high-level organization of the cell is embedded in indirect relationships that connect distinct cellular processes. Existing computational approaches for detecting indirect relationships between genes typically consist of propagating abstract information through network representations of the cell. However, the selection of genes to serve as the source of propagation is inherently biased by prior knowledge. Here, we sought to derive an unbiased view of the high-level organization of the cell by identifying the genes that propagate and receive information most effectively in the cell, and the indirect relationships between these genes. To this aim, we adapted a perturbation-response scanning strategy initially developed for identifying allosteric interactions within proteins. We deployed this strategy onto an elastic network model of the yeast genetic interaction profile similarity network. This network revealed a superior propensity for information propagation relative to simulated networks with similar topology. Perturbation-response scanning identified the major distributors and receivers of information in the network, named effector and sensor genes, respectively. Effectors formed dense clusters centrally integrated into the network, whereas sensors formed loosely connected antenna-shaped clusters and contained genes with previously characterized involvement in signal transduction. We propose that indirect relationships between effector and sensor clusters represent major paths of information flow between distinct cellular processes. Genetic similarity networks for fission yeast and human displayed similarly strong propensities for information propagation and clusters of effector and sensor genes, suggesting that the global architecture enabling indirect relationships is evolutionarily conserved across species. Our results demonstrate that elastic network modeling of cellular networks constitutes a promising strategy to probe the high-level organization and cooperativity in the cell.

Origins, evolution, and physiological implications of de novo genes in yeast.

De novo gene birth is the process by which new genes emerge in sequences that were previously noncoding. Over the past decade, researchers have taken advantage of the power of yeast as a model and a tool to study the evolutionary mechanisms and physiological implications of de novo gene birth. We summarize the mechanisms that have been proposed to explicate how noncoding sequences can become protein-coding genes, highlighting the discovery of pervasive translation of the yeast transcriptome and its presumed impact on evolutionary innovation. We summarize current best practices for the identification and characterization of de novo genes. Crucially, we explain that the field is still in its nascency, with the physiological roles of most young yeast de novo genes identified thus far still utterly unknown. We hope this review inspires researchers to investigate the true contribution of de novo gene birth to cellular physiology and phenotypic diversity across yeast strains and species.

Integrative approaches for finding modular structure in biological networks.

A central goal of systems biology is to elucidate the structural and functional architecture of the cell. To this end, large and complex networks of molecular interactions are being rapidly generated for humans and model organisms. A recent focus of bioinformatics research has been to integrate these networks with each other and with diverse molecular profiles to identify sets of molecules and interactions that participate in a common biological function - that is, 'modules'. Here, we classify such integrative approaches into four broad categories, describe their bioinformatic principles and review their applications.

Mapping transcription factor interactome networks using HaloTag protein arrays.

Protein microarrays enable investigation of diverse biochemical properties for thousands of proteins in a single experiment, an unparalleled capacity. Using a high-density system called HaloTag nucleic acid programmable protein array (HaloTag-NAPPA), we created high-density protein arrays comprising 12,000 Arabidopsis ORFs. We used these arrays to query protein-protein interactions for a set of 38 transcription factors and transcriptional regulators (TFs) that function in diverse plant hormone regulatory pathways. The resulting transcription factor interactome network, TF-NAPPA, contains thousands of novel interactions. Validation in a benchmarked in vitro pull-down assay revealed that a random subset of TF-NAPPA validated at the same rate of 64% as a positive reference set of literature-curated interactions. Moreover, using a bimolecular fluorescence complementation (BiFC) assay, we confirmed in planta several interactions of biological interest and determined the interaction localizations for seven pairs. The application of HaloTag-NAPPA technology to plant hormone signaling pathways allowed the identification of many novel transcription factor-protein interactions and led to the development of a proteome-wide plant hormone TF interactome network.

No Evidence for Phylostratigraphic Bias Impacting Inferences on Patterns of Gene Emergence and Evolution.

Phylostratigraphy is a computational framework for dating the emergence of DNA and protein sequences in a phylogeny. It has been extensively applied to make inferences on patterns of genome evolution, including patterns of disease gene evolution, ontogeny and de novo gene origination. Phylostratigraphy typically relies on BLAST searches along a species tree, but new simulation studies have raised concerns about the ability of BLAST to detect remote homologues and its impact on phylostratigraphic inferences. Here, we re-assessed these simulations. We found that, even with a possible overall BLAST false negative rate between 11-15%, the large majority of sequences assigned to a recent evolutionary origin by phylostratigraphy is unaffected by technical concerns about BLAST. Where the results of the simulations did cast doubt on previously reported findings, we repeated the original analyses but now excluded all questionable sequences. The originally described patterns remained essentially unchanged. These new analyses strongly support phylostratigraphic inferences, including: genes that emerged after the origin of eukaryotes are more likely to be expressed in the ectoderm than in the endoderm or mesoderm in Drosophila, and the de novo emergence of protein-coding genes from non-genic sequences occurs through proto-gene intermediates in yeast. We conclude that BLAST is an appropriate and sufficiently sensitive tool in phylostratigraphic analysis that does not appear to introduce significant biases into evolutionary pattern inferences.

Proto-genes and de novo gene birth.

Novel protein-coding genes can arise either through re-organization of pre-existing genes or de novo. Processes involving re-organization of pre-existing genes, notably after gene duplication, have been extensively described. In contrast, de novo gene birth remains poorly understood, mainly because translation of sequences devoid of genes, or 'non-genic' sequences, is expected to produce insignificant polypeptides rather than proteins with specific biological functions. Here we formalize an evolutionary model according to which functional genes evolve de novo through transitory proto-genes generated by widespread translational activity in non-genic sequences. Testing this model at the genome scale in Saccharomyces cerevisiae, we detect translation of hundreds of short species-specific open reading frames (ORFs) located in non-genic sequences. These translation events seem to provide adaptive potential, as suggested by their differential regulation upon stress and by signatures of retention by natural selection. In line with our model, we establish that S. cerevisiae ORFs can be placed within an evolutionary continuum ranging from non-genic sequences to genes. We identify ~1,900 candidate proto-genes among S. cerevisiae ORFs and find that de novo gene birth from such a reservoir may be more prevalent than sporadic gene duplication. Our work illustrates that evolution exploits seemingly dispensable sequences to generate adaptive functional innovation.

Interpreting cancer genomes using systematic host network perturbations by tumour virus proteins.

Genotypic differences greatly influence susceptibility and resistance to disease. Understanding genotype-phenotype relationships requires that phenotypes be viewed as manifestations of network properties, rather than simply as the result of individual genomic variations. Genome sequencing efforts have identified numerous germline mutations, and large numbers of somatic genomic alterations, associated with a predisposition to cancer. However, it remains difficult to distinguish background, or 'passenger', cancer mutations from causal, or 'driver', mutations in these data sets. Human viruses intrinsically depend on their host cell during the course of infection and can elicit pathological phenotypes similar to those arising from mutations. Here we test the hypothesis that genomic variations and tumour viruses may cause cancer through related mechanisms, by systematically examining host interactome and transcriptome network perturbations caused by DNA tumour virus proteins. The resulting integrated viral perturbation data reflects rewiring of the host cell networks, and highlights pathways, such as Notch signalling and apoptosis, that go awry in cancer. We show that systematic analyses of host targets of viral proteins can identify cancer genes with a success rate on a par with their identification through functional genomics and large-scale cataloguing of tumour mutations. Together, these complementary approaches increase the specificity of cancer gene identification. Combining systems-level studies of pathogen-encoded gene products with genomic approaches will facilitate the prioritization of cancer-causing driver genes to advance the understanding of the genetic basis of human cancer.

Genome-wide identification of Pseudomonas aeruginosa virulence-related genes using a Caenorhabditis elegans infection model.

Pseudomonas aeruginosa strain PA14 is an opportunistic human pathogen capable of infecting a wide range of organisms including the nematode Caenorhabditis elegans. We used a non-redundant transposon mutant library consisting of 5,850 clones corresponding to 75% of the total and approximately 80% of the non-essential PA14 ORFs to carry out a genome-wide screen for attenuation of PA14 virulence in C. elegans. We defined a functionally diverse 180 mutant set (representing 170 unique genes) necessary for normal levels of virulence that included both known and novel virulence factors. Seven previously uncharacterized virulence genes (ABC transporters PchH and PchI, aminopeptidase PepP, ATPase/molecular chaperone ClpA, cold shock domain protein PA0456, putative enoyl-CoA hydratase/isomerase PA0745, and putative transcriptional regulator PA14_27700) were characterized with respect to pigment production and motility and all but one of these mutants exhibited pleiotropic defects in addition to their avirulent phenotype. We examined the collection of genes required for normal levels of PA14 virulence with respect to occurrence in P. aeruginosa strain-specific genomic regions, location on putative and known genomic islands, and phylogenetic distribution across prokaryotes. Genes predominantly contributing to virulence in C. elegans showed neither a bias for strain-specific regions of the P. aeruginosa genome nor for putatively horizontally transferred genomic islands. Instead, within the collection of virulence-related PA14 genes, there was an overrepresentation of genes with a broad phylogenetic distribution that also occur with high frequency in many prokaryotic clades, suggesting that in aggregate the genes required for PA14 virulence in C. elegans are biased towards evolutionarily conserved genes.

Integrative detection of genome-wide translation using iRibo.

Ribosome profiling is a sequencing technique that provides a global picture of translation across a genome. Here, we present iRibo, a software program for integrating any number of ribosome profiling samples to obtain sensitive inference of annotated or unannotated translated open reading frames. We describe the process of using iRibo to generate a species' translatome from a set of ribosome profiling samples using S. cerevisiae as an example. For complete details on the use and execution of this protocol, please refer to Wacholder et al. (2023).1.

Biological Factors and Statistical Limitations Prevent Detection of Most Noncanonical Proteins by Mass Spectrometry.

Ribosome profiling experiments indicate pervasive translation of short open reading frames (ORFs) outside of annotated protein-coding genes. However, shotgun mass spectrometry experiments typically detect only a small fraction of the predicted protein products of this noncanonical translation. The rarity of detection could indicate that most predicted noncanonical proteins are rapidly degraded and not present in the cell; alternatively, it could reflect technical limitations. Here we leveraged recent advances in ribosome profiling and mass spectrometry to investigate the factors limiting detection of noncanonical proteins in yeast. We show that the low detection rate of noncanonical ORF products can largely be explained by small size and low translation levels and does not indicate that they are unstable or biologically insignificant. In particular, proteins encoded by evolutionarily young genes, including those with well-characterized biological roles, are too short and too lowly-expressed to be detected by shotgun mass spectrometry at current detection sensitivities. Additionally, we find that decoy biases can give misleading estimates of noncanonical protein false discovery rates, potentially leading to false detections. After accounting for these issues, we found strong evidence for four noncanonical proteins in mass spectrometry data, which were also supported by evolution and translation data. These results illustrate the power of mass spectrometry to validate unannotated genes predicted by ribosome profiling, but also its substantial limitations in finding many biologically relevant lowly-expressed proteins.

A vast evolutionarily transient translatome contributes to phenotype and fitness.

Translation is the process by which ribosomes synthesize proteins. Ribosome profiling recently revealed that many short sequences previously thought to be noncoding are pervasively translated. To identify protein-coding genes in this noncanonical translatome, we combine an integrative framework for extremely sensitive ribosome profiling analysis, iRibo, with high-powered selection inferences tailored for short sequences. We construct a reference translatome for Saccharomyces cerevisiae comprising 5,400 canonical and almost 19,000 noncanonical translated elements. Only 14 noncanonical elements were evolving under detectable purifying selection. A representative subset of translated elements lacking signatures of selection demonstrated involvement in processes including DNA repair, stress response, and post-transcriptional regulation. Our results suggest that most translated elements are not conserved protein-coding genes and contribute to genotype-phenotype relationships through fast-evolving molecular mechanisms.

Unannotated Open Reading Frame in Saccharomyces cerevisiae Encodes Protein Localizing to the Endoplasmic Reticulum.

There are thousands of unannotated translated open reading frames (ORFs) in the Saccharomyces cerevisiae genome. Previous investigation into one such unannotated ORF, which was systemically labeled YGR016C-A based on its genomic coordinates, showed that replacing the ORF's ATG start codon with AAG led to a change in cellular fitness under different stress conditions (Wacholder et al., 2023). This suggested translation of YGR016C-A plays a role in cellular fitness. Here, we investigate Ygr016c-a's subcellular localization to gain insight into its cellular function. Computational prediction tools, co-expression analysis and fluorescence microscopy suggest that the Ygr016c-a protein localizes to the endoplasmic reticulum.

Literature-curated protein interaction datasets.

High-quality datasets are needed to understand how global and local properties of protein-protein interaction, or 'interactome', networks relate to biological mechanisms, and to guide research on individual proteins. In an evaluation of existing curation of protein interaction experiments reported in the literature, we found that curation can be error-prone and possibly of lower quality than commonly assumed.

Empirically controlled mapping of the Caenorhabditis elegans protein-protein interactome network.

To provide accurate biological hypotheses and elucidate global properties of cellular networks, systematic identification of protein-protein interactions must meet high quality standards.We present an expanded C. elegans protein-protein interaction network, or 'interactome' map, derived from testing a matrix of approximately 10,000 x approximately 10,000 proteins using a highly specific, high-throughput yeast two-hybrid system. Through a new empirical quality control framework, we show that the resulting data set (Worm Interactome 2007, or WI-2007) was similar in quality to low-throughput data curated from the literature. We filtered previous interaction data sets and integrated them with WI-2007 to generate a high-confidence consolidated map (Worm Interactome version 8, or WI8). This work allowed us to estimate the size of the worm interactome at approximately 116,000 interactions. Comparison with other types of functional genomic data shows the complementarity of distinct experimental approaches in predicting different functional relationships between genes or proteins