CNV-ClinViewer: enhancing the clinical interpretation of large copy-number variants online.

MOTIVATION: Pathogenic copy-number variants (CNVs) can cause a heterogeneous spectrum of rare and severe disorders. However, most CNVs are benign and are part of natural variation in human genomes. CNV pathogenicity classification, genotype-phenotype analyses, and therapeutic target identification are challenging and time-consuming tasks that require the integration and analysis of information from multiple scattered sources by experts. RESULTS: Here, we introduce the CNV-ClinViewer, an open-source web application for clinical evaluation and visual exploration of CNVs. The application enables real-time interactive exploration of large CNV datasets in a user-friendly designed interface and facilitates semi-automated clinical CNV interpretation following the ACMG guidelines by integrating the ClassifCNV tool. In combination with clinical judgment, the application enables clinicians and researchers to formulate novel hypotheses and guide their decision-making process. Subsequently, the CNV-ClinViewer enhances for clinical investigators' patient care and for basic scientists' translational genomic research. AVAILABILITY AND IMPLEMENTATION: The web application is freely available at https://cnv-ClinViewer.broadinstitute.org and the open-source code can be found at https://github.com/LalResearchGroup/CNV-clinviewer.

Conserved patterns across ion channels correlate with variant pathogenicity and clinical phenotypes.

Clinically identified genetic variants in ion channels can be benign or cause disease by increasing or decreasing the protein function. As a consequence, therapeutic decision-making is challenging without molecular testing of each variant. Our biophysical knowledge of ion-channel structures and function is just emerging, and it is currently not well understood which amino acid residues cause disease when mutated. We sought to systematically identify biological properties associated with variant pathogenicity across all major voltage and ligand-gated ion-channel families. We collected and curated 3049 pathogenic variants from hundreds of neurodevelopmental and other disorders and 12 546 population variants for 30 ion channel or channel subunits for which a high-quality protein structure was available. Using a wide range of bioinformatics approaches, we computed 163 structural features and tested them for pathogenic variant enrichment. We developed a novel 3D spatial distance scoring approach that enables comparisons of pathogenic and population variant distribution across protein structures. We discovered and independently replicated that several pore residue properties and proximity to the pore axis were most significantly enriched for pathogenic variants compared to population variants. Using our 3D scoring approach, we showed that the strongest pathogenic variant enrichment was observed for pore-lining residues and alpha-helix residues within 5Å distance from the pore axis centre and not involved in gating. Within the subset of residues located at the pore, the hydrophobicity of the pore was the feature most strongly associated with variant pathogenicity. We also found an association between the identified properties and both clinical phenotypes and functional in vitro assays for voltage-gated sodium channels (SCN1A, SCN2A, SCN8A) and N-methyl-D-aspartate receptor (GRIN1, GRIN2A, GRIN2B) encoding genes. In an independent expert-curated dataset of 1422 neurodevelopmental disorder pathogenic patient variants and 679 electrophysiological experiments, we show that pore axis distance is associated with seizure age of onset and cognitive performance as well as differential gain versus loss-of-channel function. In summary, we identified biological properties associated with ion-channel malfunction and show that these are correlated with in vitro functional readouts and clinical phenotypes in patients with neurodevelopmental disorders. Our results suggest that clinical decision support algorithms that predict variant pathogenicity and function are feasible in the future.

SLC6A1 variant pathogenicity, molecular function and phenotype: a genetic and clinical analysis.

Genetic variants in the SLC6A1 gene can cause a broad phenotypic disease spectrum by altering the protein function. Thus, systematically curated clinically relevant genotype-phenotype associations are needed to understand the disease mechanism and improve therapeutic decision-making. We aggregated genetic and clinical data from 172 individuals with likely pathogenic/pathogenic (lp/p) SLC6A1 variants and functional data for 184 variants (14.1% lp/p). Clinical and functional data were available for a subset of 126 individuals. We explored the potential associations of variant positions on the GAT1 3D structure with variant pathogenicity, altered molecular function and phenotype severity using bioinformatic approaches. The GAT1 transmembrane domains 1, 6 and extracellular loop 4 (EL4) were enriched for patient over population variants. Across functionally tested missense variants (n = 156), the spatial proximity from the ligand was associated with loss-of-function in the GAT1 transporter activity. For variants with complete loss of in vitro GABA uptake, we found a 4.6-fold enrichment in patients having severe disease versus non-severe disease (P = 2.9 × 10-3, 95% confidence interval: 1.5-15.3). In summary, we delineated associations between the 3D structure and variant pathogenicity, variant function and phenotype in SLC6A1-related disorders. This knowledge supports biology-informed variant interpretation and research on GAT1 function. All our data can be interactively explored in the SLC6A1 portal (https://slc6a1-portal.broadinstitute.org/).

DDGun: an untrained predictor of protein stability changes upon amino acid variants.

Estimating the functional effect of single amino acid variants in proteins is fundamental for predicting the change in the thermodynamic stability, measured as the difference in the Gibbs free energy of unfolding, between the wild-type and the variant protein (ΔΔG). Here, we present the web-server of the DDGun method, which was previously developed for the ΔΔG prediction upon amino acid variants. DDGun is an untrained method based on basic features derived from evolutionary information. It is antisymmetric, as it predicts opposite ΔΔG values for direct (A → B) and reverse (B → A) single and multiple site variants. DDGun is available in two versions, one based on only sequence information and the other one based on sequence and structure information. Despite being untrained, DDGun reaches prediction performances comparable to those of trained methods. Here we make DDGun available as a web server. For the web server version, we updated the protein sequence database used for the computation of the evolutionary features, and we compiled two new data sets of protein variants to do a blind test of its performances. On these blind data sets of single and multiple site variants, DDGun confirms its prediction performance, reaching an average correlation coefficient between experimental and predicted ΔΔG of 0.45 and 0.49 for the sequence-based and structure-based versions, respectively. Besides being used for the prediction of ΔΔG, we suggest that DDGun should be adopted as a benchmark method to assess the predictive capabilities of newly developed methods. Releasing DDGun as a web-server, stand-alone program and docker image will facilitate the necessary process of method comparison to improve ΔΔG prediction.

An In Vivo Whole-Transcriptomic Approach to Assess Developmental and Reproductive Impairments Caused by Flumequine in Daphnia magna.

Among veterinary antibiotics, flumequine (FLU) is still widely used in aquaculture due to its efficacy and cost-effectiveness. Although it was synthesized more than 50 years ago, a complete toxicological framework of possible side effects on non-target species is still far from being achieved. The aim of this research was to investigate the FLU molecular mechanisms in Daphnia magna, a planktonic crustacean recognized as a model species for ecotoxicological studies. Two different FLU concentrations (2.0 mg L-1 and 0.2 mg L-1) were assayed in general accordance with OECD Guideline 211, with some proper adaptations. Exposure to FLU (2.0 mg L-1) caused alteration of phenotypic traits, with a significant reduction in survival rate, body growth, and reproduction. The lower concentration (0.2 mg L-1) did not affect phenotypic traits but modulated gene expression, an effect which was even more evident under the higher exposure level. Indeed, in daphnids exposed to 2.0 mg L-1 FLU, several genes related with growth, development, structural components, and antioxidant response were significantly modulated. To the best of our knowledge, this is the first work showing the impact of FLU on the transcriptome of D. magna.

Fido-SNP: the first webserver for scoring the impact of single nucleotide variants in the dog genome.

As the amount of genomic variation data increases, tools that are able to score the functional impact of single nucleotide variants become more and more necessary. While there are several prediction servers available for interpreting the effects of variants in the human genome, only few have been developed for other species, and none were specifically designed for species of veterinary interest such as the dog. Here, we present Fido-SNP the first predictor able to discriminate between Pathogenic and Benign single-nucleotide variants in the dog genome. Fido-SNP is a binary classifier based on the Gradient Boosting algorithm. It is able to classify and score the impact of variants in both coding and non-coding regions based on sequence features within seconds. When validated on a previously unseen set of annotated variants from the OMIA database, Fido-SNP reaches 88% overall accuracy, 0.77 Matthews correlation coefficient and 0.91 Area Under the ROC Curve.

A natural upper bound to the accuracy of predicting protein stability changes upon mutations.

MOTIVATION: Accurate prediction of protein stability changes upon single-site variations (ΔΔG) is important for protein design, as well as for our understanding of the mechanisms of genetic diseases. The performance of high-throughput computational methods to this end is evaluated mostly based on the Pearson correlation coefficient between predicted and observed data, assuming that the upper bound would be 1 (perfect correlation). However, the performance of these predictors can be limited by the distribution and noise of the experimental data. Here we estimate, for the first time, a theoretical upper-bound to the ΔΔG prediction performances imposed by the intrinsic structure of currently available ΔΔG data. RESULTS: Given a set of measured ΔΔG protein variations, the theoretically "best predictor" is estimated based on its similarity to another set of experimentally determined ΔΔG values. We investigate the correlation between pairs of measured ΔΔG variations, where one is used as a predictor for the other. We analytically derive an upper bound to the Pearson correlation as a function of the noise and distribution of the ΔΔG data. We also evaluate the available datasets to highlight the effect of the noise in conjunction with ΔΔG distribution. We conclude that the upper bound is a function of both uncertainty and spread of the ΔΔG values, and that with current data the best performance should be between 0.7 and 0.8, depending on the dataset used; higher Pearson correlations might be indicative of overtraining. It also follows that comparisons of predictors using different datasets are inherently misleading. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

DDGun: an untrained method for the prediction of protein stability changes upon single and multiple point variations.

BACKGROUND: Predicting the effect of single point variations on protein stability constitutes a crucial step toward understanding the relationship between protein structure and function. To this end, several methods have been developed to predict changes in the Gibbs free energy of unfolding (∆∆G) between wild type and variant proteins, using sequence and structure information. Most of the available methods however do not exhibit the anti-symmetric prediction property, which guarantees that the predicted ∆∆G value for a variation is the exact opposite of that predicted for the reverse variation, i.e., ∆∆G(A → B) = -∆∆G(B → A), where A and B are amino acids. RESULTS: Here we introduce simple anti-symmetric features, based on evolutionary information, which are combined to define an untrained method, DDGun (DDG untrained). DDGun is a simple approach based on evolutionary information that predicts the ∆∆G for single and multiple variations from sequence and structure information (DDGun3D). Our method achieves remarkable performance without any training on the experimental datasets, reaching Pearson correlation coefficients between predicted and measured ∆∆G values of ~ 0.5 and ~ 0.4 for single and multiple site variations, respectively. Surprisingly, DDGun performances are comparable with those of state of the art methods. DDGun also naturally predicts multiple site variations, thereby defining a benchmark method for both single site and multiple site predictors. DDGun is anti-symmetric by construction predicting the value of the ∆∆G of a reciprocal variation as almost equal (depending on the sequence profile) to -∆∆G of the direct variation. This is a valuable property that is missing in the majority of the methods. CONCLUSIONS: Evolutionary information alone combined in an untrained method can achieve remarkably high performances in the prediction of ∆∆G upon protein mutation. Non-trained approaches like DDGun represent a valid benchmark both for scoring the predictive power of the individual features and for assessing the learning capability of supervised methods.

Dynamic sensitivity and nonlinear interactions influence the system-level evolutionary patterns of phototransduction proteins.

Determining the influence of complex, molecular-system dynamics on the evolution of proteins is hindered by the significant challenge of quantifying the control exerted by the proteins on system output. We have employed a combination of systems biology and molecular evolution analyses in a first attempt to unravel this relationship. We employed a comprehensive mathematical model of mammalian phototransduction to predict the degree of influence that each protein in the system exerts on the high-level dynamic behaviour. We found that the genes encoding the most dynamically sensitive proteins exhibit relatively relaxed evolutionary constraint. We also investigated the evolutionary and epistatic influences of the many nonlinear interactions between proteins in the system and found several pairs to have coevolved, including those whose interactions are purely dynamical with respect to system output. This evidence points to a key role played by nonlinear system dynamics in influencing patterns of molecular evolution.

Distances from ligands as main predictive features for pathogenicity and functional effect of variants in NMDA receptors.

Genetic variants in genes GRIN1 , GRIN2A , GRIN2B , and GRIN2D , which encode subunits of the N-methyl-D-aspartate receptor (NMDAR), have been associated with severe and heterogeneous neurologic diseases. Missense variants in these genes can result in gain or loss of the NMDAR function, requiring opposite therapeutic treatments. Computational methods that predict pathogenicity and molecular functional effects are therefore crucial for accurate diagnosis and therapeutic applications. We assembled missense variants: 201 from patients, 631 from general population, and 159 characterized by electrophysiological readouts showing whether they can enhance or reduce the receptor function. This includes new functional data from 47 variants reported here, for the first time. We found that pathogenic/benign variants and variants that increase/decrease the channel function were distributed unevenly on the protein structure, with spatial proximity to ligands bound to the agonist and antagonist binding sites being key predictive features. Leveraging distances from ligands, we developed two independent machine learning-based predictors for NMDAR missense variants: a pathogenicity predictor which outperforms currently available predictors (AUC=0.945, MCC=0.726), and the first binary predictor of molecular function (increase or decrease) (AUC=0.809, MCC=0.523). Using these, we reclassified variants of uncertain significance in the ClinVar database and refined a previous genome-informed epidemiological model to estimate the birth incidence of molecular mechanism-defined GRIN disorders. Our findings demonstrate that distance from ligands is an important feature in NMDARs that can enhance variant pathogenicity prediction and enable functional prediction. Further studies with larger numbers of phenotypically and functionally characterized variants will enhance the potential clinical utility of this method.

Generation and characterization of cytochrome P450 3A74 CRISPR/Cas9 knockout bovine foetal hepatocyte cell line (BFH12).

In human, the cytochrome P450 3A (CYP3A) subfamily of drug-metabolizing enzymes (DMEs) is responsible for a significant number of phase I reactions, with the CYP3A4 isoform superintending the hepatic and intestinal metabolism of diverse endobiotic and xenobiotic compounds. The CYP3A4-dependent bioactivation of chemicals may result in hepatotoxicity and trigger carcinogenesis. In cattle, four CYP3A genes (CYP3A74, CYP3A76, CYP3A28 and CYP3A24) have been identified. Despite cattle being daily exposed to xenobiotics (e.g., mycotoxins, food additives, drugs and pesticides), the existing knowledge about the contribution of CYP3A in bovine hepatic metabolism is still incomplete. Nowadays, CRISPR/Cas9 mediated knockout (KO) is a valuable method to generate in vivo and in vitro models for studying the metabolism of xenobiotics. In the present study, we successfully performed CRISPR/Cas9-mediated KO of bovine CYP3A74, human CYP3A4-like, in a bovine foetal hepatocyte cell line (BFH12). After clonal expansion and selection, CYP3A74 ablation was confirmed at the DNA, mRNA, and protein level. The subsequent characterization of the CYP3A74 KO clone highlighted significant transcriptomic changes (RNA-sequencing) associated with the regulation of cell cycle and proliferation, immune and inflammatory response, as well as metabolic processes. Overall, this study successfully developed a new CYP3A74 KO in vitro model by using CRISPR/Cas9 technology, which represents a novel resource for xenobiotic metabolism studies in cattle. Furthermore, the transcriptomic analysis suggests a key role of CYP3A74 in bovine hepatocyte cell cycle regulation and metabolic homeostasis.

A system-level, molecular evolutionary analysis of mammalian phototransduction.

BACKGROUND: Visual perception is initiated in the photoreceptor cells of the retina via the phototransduction system. This system has shown marked evolution during mammalian divergence in such complex attributes as activation time and recovery time. We have performed a molecular evolutionary analysis of proteins involved in mammalian phototransduction in order to unravel how the action of natural selection has been distributed throughout the system to evolve such traits. RESULTS: We found selective pressures to be non-randomly distributed according to both a simple protein classification scheme and a protein-interaction network representation of the signaling pathway. Proteins which are topologically central in the signaling pathway, such as the G proteins, as well as retinoid cycle chaperones and proteins involved in photoreceptor cell-type determination, were found to be more constrained in their evolution. Proteins peripheral to the pathway, such as ion channels and exchangers, as well as the retinoid cycle enzymes, have experienced a relaxation of selective pressures. Furthermore, signals of positive selection were detected in two genes: the short-wave (blue) opsin (OPN1SW) in hominids and the rod-specific Na+/ Ca2+, K+ ion exchanger (SLC24A1) in rodents. CONCLUSIONS: The functions of the proteins involved in phototransduction and the topology of the interactions between them have imposed non-random constraints on their evolution. Thus, in shaping or conserving system-level phototransduction traits, natural selection has targeted the underlying proteins in a concerted manner.

Exploring the rate-limiting steps in visual phototransduction recovery by bottom-up kinetic modeling.

BACKGROUND: Phototransduction in vertebrate photoreceptor cells represents a paradigm of signaling pathways mediated by G-protein-coupled receptors (GPCRs), which share common modules linking the initiation of the cascade to the final response of the cell. In this work, we focused on the recovery phase of the visual photoresponse, which is comprised of several interacting mechanisms. RESULTS: We employed current biochemical knowledge to investigate the response mechanisms of a comprehensive model of the visual phototransduction pathway. In particular, we have improved the model by implementing a more detailed representation of the recoverin (Rec)-mediated calcium feedback on rhodopsin kinase and including a dynamic arrestin (Arr) oligomerization mechanism. The model was successfully employed to investigate the rate limiting steps in the recovery of the rod photoreceptor cell after illumination. Simulation of experimental conditions in which the expression levels of rhodospin kinase (RK), of the regulator of the G-protein signaling (RGS), of Arr and of Rec were altered individually or in combination revealed severe kinetic constraints to the dynamics of the overall network. CONCLUSIONS: Our simulations confirm that RGS-mediated effector shutdown is the rate-limiting step in the recovery of the photoreceptor and show that the dynamic formation and dissociation of Arr homodimers and homotetramers at different light intensities significantly affect the timing of rhodopsin shutdown. The transition of Arr from its oligomeric storage forms to its monomeric form serves to temper its availability in the functional state. Our results may explain the puzzling evidence that overexpressing RK does not influence the saturation time of rod cells at bright light stimuli. The approach presented here could be extended to the study of other GPCR signaling pathways.

Discovering the Protective Effects of Quercetin on Aflatoxin B1-Induced Toxicity in Bovine Foetal Hepatocyte-Derived Cells (BFH12).

Aflatoxin B1 (AFB1) induces lipid peroxidation and mortality in bovine foetal hepatocyte-derived cells (BFH12), with underlying transcriptional perturbations associated mainly with cancer, cellular damage, inflammation, bioactivation, and detoxification pathways. In this cell line, curcumin and resveratrol have proven to be effective in mitigating AFB1-induced toxicity. In this paper, we preliminarily assessed the potential anti-AFB1 activity of a natural polyphenol, quercetin (QUE), in BFH12 cells. To this end, we primarily measured QUE cytotoxicity using a WST-1 reagent. Then, we pre-treated the cells with QUE and exposed them to AFB1. The protective role of QUE was evaluated by measuring cytotoxicity, transcriptional changes (RNA-sequencing), lipid peroxidation (malondialdehyde production), and targeted post-transcriptional modifications (NQO1 and CYP3A enzymatic activity). The results demonstrated that QUE, like curcumin and resveratrol, reduced AFB1-induced cytotoxicity and lipid peroxidation and caused larger transcriptional variations than AFB1 alone. Most of the differentially expressed genes were involved in lipid homeostasis, inflammatory and immune processes, and carcinogenesis. As for enzymatic activities, QUE significantly reverted CYP3A variations induced by AFB1, but not those of NQO1. This study provides new knowledge about key molecular mechanisms involved in QUE-mediated protection against AFB1 toxicity and encourages in vivo studies to assess QUE's bioavailability and beneficial effects on aflatoxicosis.

Genome-wide identification and phenotypic characterization of seizure-associated copy number variations in 741,075 individuals.

Copy number variants (CNV) are established risk factors for neurodevelopmental disorders with seizures or epilepsy. With the hypothesis that seizure disorders share genetic risk factors, we pooled CNV data from 10,590 individuals with seizure disorders, 16,109 individuals with clinically validated epilepsy, and 492,324 population controls and identified 25 genome-wide significant loci, 22 of which are novel for seizure disorders, such as deletions at 1p36.33, 1q44, 2p21-p16.3, 3q29, 8p23.3-p23.2, 9p24.3, 10q26.3, 15q11.2, 15q12-q13.1, 16p12.2, 17q21.31, duplications at 2q13, 9q34.3, 16p13.3, 17q12, 19p13.3, 20q13.33, and reciprocal CNVs at 16p11.2, and 22q11.21. Using genetic data from additional 248,751 individuals with 23 neuropsychiatric phenotypes, we explored the pleiotropy of these 25 loci. Finally, in a subset of individuals with epilepsy and detailed clinical data available, we performed phenome-wide association analyses between individual CNVs and clinical annotations categorized through the Human Phenotype Ontology (HPO). For six CNVs, we identified 19 significant associations with specific HPO terms and generated, for all CNVs, phenotype signatures across 17 clinical categories relevant for epileptologists. This is the most comprehensive investigation of CNVs in epilepsy and related seizure disorders, with potential implications for clinical practice.

Distribution of events of positive selection and population differentiation in a metabolic pathway: the case of asparagine N-glycosylation.

BACKGROUND: Asparagine N-Glycosylation is one of the most important forms of protein post-translational modification in eukaryotes. This metabolic pathway can be subdivided into two parts: an upstream sub-pathway required for achieving proper folding for most of the proteins synthesized in the secretory pathway, and a downstream sub-pathway required to give variability to trans-membrane proteins, and involved in adaptation to the environment and innate immunity. Here we analyze the nucleotide variability of the genes of this pathway in human populations, identifying which genes show greater population differentiation and which genes show signatures of recent positive selection. We also compare how these signals are distributed between the upstream and the downstream parts of the pathway, with the aim of exploring how forces of population differentiation and positive selection vary among genes involved in the same metabolic pathway but subject to different functional constraints. RESULTS: Our results show that genes in the downstream part of the pathway are more likely to show a signature of population differentiation, while events of positive selection are equally distributed among the two parts of the pathway. Moreover, events of positive selection are frequent on genes that are known to be at bifurcation points, and that are identified as being in key position by a network-level analysis such as MGAT3 and GCS1. CONCLUSIONS: These findings indicate that the upstream part of the Asparagine N-Glycosylation pathway has lower diversity among populations, while the downstream part is freer to tolerate diversity among populations. Moreover, the distribution of signatures of population differentiation and positive selection can change between parts of a pathway, especially between parts that are exposed to different functional constraints. Our results support the hypothesis that genes involved in constitutive processes can be expected to show lower population differentiation, while genes involved in traits related to the environment should show higher variability. Taken together, this work broadens our knowledge on how events of population differentiation and of positive selection are distributed among different parts of a metabolic pathway.

Molecular evolution and network-level analysis of the N-glycosylation metabolic pathway across primates.

N-glycosylation is one of the most important forms of protein modification, serving key biological functions in multicellular organisms. N-glycans at the cell surface mediate the interaction between cells and the surrounding matrix and may act as pathogen receptors, making the genes responsible for their synthesis good candidates to show signatures of adaptation to different pathogen environments. Here, we study the forces that shaped the evolution of the genes involved in the synthesis of the N-glycans during the divergence of primates within the framework of their functional network. We have found that, despite their function of producing glycan repertoires capable of evading rapidly evolving pathogens, genes involved in the synthesis of the glycans are highly conserved, and no signals of positive selection have been detected within the time of divergence of primates. This suggests strong functional constraints as the main force driving their evolution. We studied the strength of the purifying selection acting on the genes in relation to the network structure considering the position of each gene along the pathway, its connectivity, and the rates of evolution in neighboring genes. We found a strong and highly significant negative correlation between the strength of purifying selection and the connectivity of each gene, indicating that genes encoding for highly connected enzymes evolve slower and thus are subject to stronger selective constraints. This result confirms that network topology does shape the evolution of the genes and that the connectivity within metabolic pathways and networks plays a major role in constraining evolutionary rates.

Genetic adaptation of the antibacterial human innate immunity network.

BACKGROUND: Pathogens have represented an important selective force during the adaptation of modern human populations to changing social and other environmental conditions. The evolution of the immune system has therefore been influenced by these pressures. Genomic scans have revealed that immune system is one of the functions enriched with genes under adaptive selection. RESULTS: Here, we describe how the innate immune system has responded to these challenges, through the analysis of resequencing data for 132 innate immunity genes in two human populations. Results are interpreted in the context of the functional and interaction networks defined by these genes. Nucleotide diversity is lower in the adaptors and modulators functional classes, and is negatively correlated with the centrality of the proteins within the interaction network. We also produced a list of candidate genes under positive or balancing selection in each population detected by neutrality tests and showed that some functional classes are preferential targets for selection. CONCLUSIONS: We found evidence that the role of each gene in the network conditions the capacity to evolve or their evolvability: genes at the core of the network are more constrained, while adaptation mostly occurred at particular positions at the network edges. Interestingly, the functional classes containing most of the genes with signatures of balancing selection are involved in autoinflammatory and autoimmune diseases, suggesting a counterbalance between the beneficial and deleterious effects of the immune response.