Proteomic prediction of diverse incident diseases: a machine learning-guided biomarker discovery study using data from a prospective cohort study.

BACKGROUND: Broad-capture proteomic technologies have the potential to improve disease prediction, enabling targeted prevention and management, but studies have so far been limited to very few selected diseases and have not evaluated predictive performance across multiple conditions. We aimed to evaluate the potential of serum proteins to improve risk prediction over and above health-derived information and polygenic risk scores across a diverse set of 24 outcomes. METHODS: We designed multiple case-cohorts nested in the EPIC-Norfolk prospective study, from participants with available serum samples and genome-wide genotype data, with more than 32 974 person-years of follow-up. Participants were middle-aged individuals (aged 40-79 years at baseline) of European ancestry who were recruited from the general population of Norfolk, England, between March, 1993 and December, 1997. We selected participants who developed one of ten less common diseases within 10 years of follow-up; we also subsampled a randomly drawn control subcohort, which also served to investigate 14 more common outcomes (n>70), including all-cause premature mortality (death before the age of 75 years; case numbers 71-437; controls 608-1556). Individuals were excluded from the current study owing to failed genotyping or proteomic quality control, relatedness, or missing information on age, sex, BMI, or smoking status. We used a machine learning framework to derive sparse predictive protein models for the onset of the the 23 individual diseases and all-cause premature mortality, and to derive a single common sparse multimorbidity signature that was predictive across multiple diseases from 2923 serum proteins. FINDINGS: Participants who developed one of ten less common diseases within 10 years of follow-up included 482 women and 507 men, with a mean age at baseline of 64·56 years (8·08). The random subcohort included 990 women and 769 men, with a mean age of 58·79 years (9·31). As few as five proteins alone outperformed polygenic risk scores for 17 of 23 outcomes (median dfference in concordance index [C-index] 0·13 [0·10-0·17]) and improved predictive performance when added over basic patient-derived information models for seven outcomes, achieving a median C-index of 0·82 (IQR 0·77-0·82). This included diseases with poor prognosis such as lung cancer (C-index 0·85 [+/- cross-validation error 0·83-0·87]), for which we identified unreported biomarkers such as C-X-C motif chemokine ligand 17. A sparse multimorbidity signature of ten proteins improved prediction across seven outcomes over patient-derived information models, achieving performances (median C-index 0·81 [IQR 0·80-0·82]) similar to those of disease-specific signatures. INTERPRETATION: We show the value of broad-capture proteomic biomarker discovery studies across multiple diseases of diverse causes, pointing to those that might benefit the most from proteomic approaches, and the potential to derive common sparse biomarker panels for prediction of multiple diseases at once. This framework could enable follow-up studies to explore the generalisability of proteomic models and to benchmark these against clinical assays, which are required to understand the translational potential of these findings. FUNDING: Medical Research Council, Health Data Research UK, UK Research and Innovation-National Institute for Health and Care Research, Cancer Research UK, and Wellcome Trust.

Multi-omic prediction of incident type 2 diabetes.

AIMS/HYPOTHESIS: The identification of people who are at high risk of developing type 2 diabetes is a key part of population-level prevention strategies. Previous studies have evaluated the predictive utility of omics measurements, such as metabolites, proteins or polygenic scores, but have considered these separately. The improvement that combined omics biomarkers can provide over and above current clinical standard models is unclear. The aim of this study was to test the predictive performance of genome, proteome, metabolome and clinical biomarkers when added to established clinical prediction models for type 2 diabetes. METHODS: We developed sparse interpretable prediction models in a prospective, nested type 2 diabetes case-cohort study (N=1105, incident type 2 diabetes cases=375) with 10,792 person-years of follow-up, selecting from 5759 features across the genome, proteome, metabolome and clinical biomarkers using least absolute shrinkage and selection operator (LASSO) regression. We compared the predictive performance of omics-derived predictors with a clinical model including the variables from the Cambridge Diabetes Risk Score and HbA1c. RESULTS: Among single omics prediction models that did not include clinical risk factors, the top ten proteins alone achieved the highest performance (concordance index [C index]=0.82 [95% CI 0.75, 0.88]), suggesting the proteome as the most informative single omic layer in the absence of clinical information. However, the largest improvement in prediction of type 2 diabetes incidence over and above the clinical model was achieved by the top ten features across several omic layers (C index=0.87 [95% CI 0.82, 0.92], Δ C index=0.05, p=0.045). This improvement by the top ten omic features was also evident in individuals with HbA1c <42 mmol/mol (6.0%), the threshold for prediabetes (C index=0.84 [95% CI 0.77, 0.90], Δ C index=0.07, p=0.03), the group in whom prediction would be most useful since they are not targeted for preventative interventions by current clinical guidelines. In this subgroup, the type 2 diabetes polygenic risk score was the major contributor to the improvement in prediction, and achieved a comparable improvement in performance when added onto the clinical model alone (C index=0.83 [95% CI 0.75, 0.90], Δ C index=0.06, p=0.002). However, compared with those with prediabetes, individuals at high polygenic risk in this group had only around half the absolute risk for type 2 diabetes over a 20 year period. CONCLUSIONS/INTERPRETATION: Omic approaches provided marginal improvements in prediction of incident type 2 diabetes. However, while a polygenic risk score does improve prediction in people with an HbA1c in the normoglycaemic range, the group in whom prediction would be most useful, even individuals with a high polygenic burden in that subgroup had a low absolute type 2 diabetes risk. This suggests a limited feasibility of implementing targeted population-based genetic screening for preventative interventions.

Proteomic analysis of 92 circulating proteins and their effects in cardiometabolic diseases.

BACKGROUND: Human plasma contains a wide variety of circulating proteins. These proteins can be important clinical biomarkers in disease and also possible drug targets. Large scale genomics studies of circulating proteins can identify genetic variants that lead to relative protein abundance. METHODS: We conducted a meta-analysis on genome-wide association studies of autosomal chromosomes in 22,997 individuals of primarily European ancestry across 12 cohorts to identify protein quantitative trait loci (pQTL) for 92 cardiometabolic associated plasma proteins. RESULTS: We identified 503 (337 cis and 166 trans) conditionally independent pQTLs, including several novel variants not reported in the literature. We conducted a sex-stratified analysis and found that 118 (23.5%) of pQTLs demonstrated heterogeneity between sexes. The direction of effect was preserved but there were differences in effect size and significance. Additionally, we annotate trans-pQTLs with nearest genes and report plausible biological relationships. Using Mendelian randomization, we identified causal associations for 18 proteins across 19 phenotypes, of which 10 have additional genetic colocalization evidence. We highlight proteins associated with a constellation of cardiometabolic traits including angiopoietin-related protein 7 (ANGPTL7) and Semaphorin 3F (SEMA3F). CONCLUSION: Through large-scale analysis of protein quantitative trait loci, we provide a comprehensive overview of common variants associated with plasma proteins. We highlight possible biological relationships which may serve as a basis for further investigation into possible causal roles in cardiometabolic diseases.

Circulating proteins to predict COVID-19 severity.

Predicting COVID-19 severity is difficult, and the biological pathways involved are not fully understood. To approach this problem, we measured 4701 circulating human protein abundances in two independent cohorts totaling 986 individuals. We then trained prediction models including protein abundances and clinical risk factors to predict COVID-19 severity in 417 subjects and tested these models in a separate cohort of 569 individuals. For severe COVID-19, a baseline model including age and sex provided an area under the receiver operator curve (AUC) of 65% in the test cohort. Selecting 92 proteins from the 4701 unique protein abundances improved the AUC to 88% in the training cohort, which remained relatively stable in the testing cohort at 86%, suggesting good generalizability. Proteins selected from different COVID-19 severity were enriched for cytokine and cytokine receptors, but more than half of the enriched pathways were not immune-related. Taken together, these findings suggest that circulating proteins measured at early stages of disease progression are reasonably accurate predictors of COVID-19 severity. Further research is needed to understand how to incorporate protein measurement into clinical care.

Proteogenomic links to human metabolic diseases.

Studying the plasma proteome as the intermediate layer between the genome and the phenome has the potential to identify new disease processes. Here, we conducted a cis-focused proteogenomic analysis of 2,923 plasma proteins measured in 1,180 individuals using antibody-based assays. We (1) identify 256 unreported protein quantitative trait loci (pQTL); (2) demonstrate shared genetic regulation of 224 cis-pQTLs with 575 specific health outcomes, revealing examples for notable metabolic diseases (such as gastrin-releasing peptide as a potential therapeutic target for type 2 diabetes); (3) improve causal gene assignment at 40% (n = 192) of overlapping risk loci; and (4) observe convergence of phenotypic consequences of cis-pQTLs and rare loss-of-function gene burden for 12 proteins, such as TIMD4 for lipoprotein metabolism. Our findings demonstrate the value of integrating complementary proteomic technologies with genomics even at moderate scale to identify new mediators of metabolic diseases with the potential for therapeutic interventions.

Proteomic signatures for identification of impaired glucose tolerance.

The implementation of recommendations for type 2 diabetes (T2D) screening and diagnosis focuses on the measurement of glycated hemoglobin (HbA1c) and fasting glucose. This approach leaves a large number of individuals with isolated impaired glucose tolerance (iIGT), who are only detectable through oral glucose tolerance tests (OGTTs), at risk of diabetes and its severe complications. We applied machine learning to the proteomic profiles of a single fasted sample from 11,546 participants of the Fenland study to test discrimination of iIGT defined using the gold-standard OGTTs. We observed significantly improved discriminative performance by adding only three proteins (RTN4R, CBPM and GHR) to the best clinical model (AUROC = 0.80 (95% confidence interval: 0.79-0.86), P = 0.004), which we validated in an external cohort. Increased plasma levels of these candidate proteins were associated with an increased risk for future T2D in an independent cohort and were also increased in individuals genetically susceptible to impaired glucose homeostasis and T2D. Assessment of a limited number of proteins can identify individuals likely to be missed by current diagnostic strategies and at high risk of T2D and its complications.

The dynamic changes and sex differences of 147 immune-related proteins during acute COVID-19 in 580 individuals.

INTRODUCTION: Severe COVID-19 leads to important changes in circulating immune-related proteins. To date it has been difficult to understand their temporal relationship and identify cytokines that are drivers of severe COVID-19 outcomes and underlie differences in outcomes between sexes. Here, we measured 147 immune-related proteins during acute COVID-19 to investigate these questions. METHODS: We measured circulating protein abundances using the SOMAscan nucleic acid aptamer panel in two large independent hospital-based COVID-19 cohorts in Canada and the United States. We fit generalized additive models with cubic splines from the start of symptom onset to identify protein levels over the first 14 days of infection which were different between severe cases and controls, adjusting for age and sex. Severe cases were defined as individuals with COVID-19 requiring invasive or non-invasive mechanical respiratory support. RESULTS: 580 individuals were included in the analysis. Mean subject age was 64.3 (sd 18.1), and 47% were male. Of the 147 proteins, 69 showed a significant difference between cases and controls (p < 3.4 × 10-4). Three clusters were formed by 108 highly correlated proteins that replicated in both cohorts, making it difficult to determine which proteins have a true causal effect on severe COVID-19. Six proteins showed sex differences in levels over time, of which 3 were also associated with severe COVID-19: CCL26, IL1RL2, and IL3RA, providing insights to better understand the marked differences in outcomes by sex. CONCLUSIONS: Severe COVID-19 is associated with large changes in 69 immune-related proteins. Further, five proteins were associated with sex differences in outcomes. These results provide direct insights into immune-related proteins that are strongly influenced by severe COVID-19 infection.

Synergistic insights into human health from aptamer- and antibody-based proteomic profiling.

Affinity-based proteomics has enabled scalable quantification of thousands of protein targets in blood enhancing biomarker discovery, understanding of disease mechanisms, and genetic evaluation of drug targets in humans through protein quantitative trait loci (pQTLs). Here, we integrate two partly complementary techniques-the aptamer-based SomaScan® v4 assay and the antibody-based Olink assays-to systematically assess phenotypic consequences of hundreds of pQTLs discovered for 871 protein targets across both platforms. We create a genetically anchored cross-platform proteome-phenome network comprising 547 protein-phenotype connections, 36.3% of which were only seen with one of the two platforms suggesting that both techniques capture distinct aspects of protein biology. We further highlight discordance of genetically predicted effect directions between assays, such as for PILRA and Alzheimer's disease. Our results showcase the synergistic nature of these technologies to better understand and identify disease mechanisms and provide a benchmark for future cross-platform discoveries.

Mapping the proteo-genomic convergence of human diseases.

Characterization of the genetic regulation of proteins is essential for understanding disease etiology and developing therapies. We identified 10,674 genetic associations for 3892 plasma proteins to create a cis-anchored gene-protein-disease map of 1859 connections that highlights strong cross-disease biological convergence. This proteo-genomic map provides a framework to connect etiologically related diseases, to provide biological context for new or emerging disorders, and to integrate different biological domains to establish mechanisms for known gene-disease links. Our results identify proteo-genomic connections within and between diseases and establish the value of cis-protein variants for annotation of likely causal disease genes at loci identified in genome-wide association studies, thereby addressing a major barrier to experimental validation and clinical translation of genetic discoveries.

Soluble fms-like tyrosine kinase to placental growth factor ratio in different stages of early-onset fetal growth restriction and small for gestational age.

INTRODUCTION: Increased soluble fms-like tyrosine kinase to placental growth factor ratio (sFlt-1/PlGF) has been demonstrated in early-onset fetal growth restriction (FGR) and small for gestational age (SGA). sFlt-1/PlGF cut-offs have been described to assess preeclampsia severity; however, sFlt-1/PlGF values present in early-onset SGA and different FGR severity stages remain unknown. Hence, the objective of this study was to describe and compare the sFlt-1/PlGF values and pregnancy outcomes among early-onset SGA/FGR stages. MATERIAL AND METHODS: This is a prospective case-control study conducted at Vall d'Hebron University Hospital. Singleton pregnancies with estimated fetal weight <10th centile and a control group of uncomplicated pregnancies between 20+0 and 31+6  weeks of gestation were enrolled. Study women were classified at diagnosis into different stages, according to estimated fetal weight centile and Doppler ultrasound. sFlt-1/PlGF serum concentrations were measured at diagnosis and, together with pregnancy outcomes, were compared among FGR severity stages, SGA, and controls. Finally, correlations between sFlt-1/PlGF values and time to delivery, gestational age at delivery, days of neonatal admission, and birthweight z-scores were investigated. RESULTS: Among the 207 women enrolled, 32 (15.4%) had uncomplicated pregnancies, 49 (23.7%) pregnancies showed SGA, and 126 (60.9%) involved FGR (92 being stage I, 17 stage II, and 17 stage III). SGA and controls had similar median sFlt-1/PlGF values (25.7 vs 27.1, P > .05) and pregnancy outcomes. However, all FGR stages had significantly poorer outcomes and greater sFlt-1/PlGF values than those of SGA and controls. Furthermore, median values differed significantly among all FGR severity stages (9.76 for stage I; 284.3 for stage II, and 625.02 for stage III, P < .05) increasing with FGR severity as well as the frequency of adverse pregnancy outcomes. Additionally, a significant correlation was found between greater sFlt-1/PlGF ratio values and gestational age at delivery, time from diagnosis to delivery, birthweight z-scores, and time in neonatal intensive care unit (r = -.637, r = -.576, r = -.161, and r = .311, respectively). CONCLUSIONS: Values of sFlt-1/PlGF at diagnosis permit early-onset FGR/SGA severity classification with good correlation with Doppler ultrasound findings and the occurrence of adverse outcomes. Thus, sFlt-1/PlGF could aid in early-onset FGR/SGA severity classification and clinical management when Doppler assessment is not feasible.

Genetic architecture of host proteins involved in SARS-CoV-2 infection.

Understanding the genetic architecture of host proteins interacting with SARS-CoV-2 or mediating the maladaptive host response to COVID-19 can help to identify new or repurpose existing drugs targeting those proteins. We present a genetic discovery study of 179 such host proteins among 10,708 individuals using an aptamer-based technique. We identify 220 host DNA sequence variants acting in cis (MAF 0.01-49.9%) and explaining 0.3-70.9% of the variance of 97 of these proteins, including 45 with no previously known protein quantitative trait loci (pQTL) and 38 encoding current drug targets. Systematic characterization of pQTLs across the phenome identified protein-drug-disease links and evidence that putative viral interaction partners such as MARK3 affect immune response. Our results accelerate the evaluation and prioritization of new drug development programmes and repurposing of trials to prevent, treat or reduce adverse outcomes. Rapid sharing and detailed interrogation of results is facilitated through an interactive webserver ( https://omicscience.org/apps/covidpgwas/ ).

Integrating Genetics and the Plasma Proteome to Predict the Risk of Type 2 Diabetes.

PURPOSE OF THE REVIEW: Proteins are the central layer of information transfer from genome to phenome and represent the largest class of drug targets. We review recent advances in high-throughput technologies that provide comprehensive, scalable profiling of the plasma proteome with the potential to improve prediction and mechanistic understanding of type 2 diabetes (T2D). RECENT FINDINGS: Technological and analytical advancements have enabled identification of novel protein biomarkers and signatures that help to address challenges of existing approaches to predict and screen for T2D. Genetic studies have so far revealed putative causal roles for only few of the proteins that have been linked to T2D, but ongoing large-scale genetic studies of the plasma proteome will help to address this and increase our understanding of aetiological pathways and mechanisms leading to diabetes. Studies of the human plasma proteome have started to elucidate its potential for T2D prediction and biomarker discovery. Future studies integrating genomic and proteomic data will provide opportunities to prioritise drug targets and identify pathways linking genetic predisposition to T2D development.

Genetic architecture of host proteins interacting with SARS-CoV-2.

Strategies to develop therapeutics for SARS-CoV-2 infection may be informed by experimental identification of viral-host protein interactions in cellular assays and measurement of host response proteins in COVID-19 patients. Identification of genetic variants that influence the level or activity of these proteins in the host could enable rapid 'in silico' assessment in human genetic studies of their causal relevance as molecular targets for new or repurposed drugs to treat COVID-19. We integrated large-scale genomic and aptamer-based plasma proteomic data from 10,708 individuals to characterize the genetic architecture of 179 host proteins reported to interact with SARS-CoV-2 proteins or to participate in the host response to COVID-19. We identified 220 host DNA sequence variants acting in cis (MAF 0.01-49.9%) and explaining 0.3-70.9% of the variance of 97 of these proteins, including 45 with no previously known protein quantitative trait loci (pQTL) and 38 encoding current drug targets. Systematic characterization of pQTLs across the phenome identified protein-drug-disease links, evidence that putative viral interaction partners such as MARK3 affect immune response, and establish the first link between a recently reported variant for respiratory failure of COVID-19 patients at the ABO locus and hypercoagulation, i.e. maladaptive host response. Our results accelerate the evaluation and prioritization of new drug development programmes and repurposing of trials to prevent, treat or reduce adverse outcomes. Rapid sharing and dynamic and detailed interrogation of results is facilitated through an interactive webserver ( https://omicscience.org/apps/covidpgwas/ ).

Voltage-dependent gating and gating charge measurements in the Kv1.2 potassium channel.

Much has been learned about the voltage sensors of ion channels since the x-ray structure of the mammalian voltage-gated potassium channel Kv1.2 was published in 2005. High resolution structural data of a Kv channel enabled the structural interpretation of numerous electrophysiological findings collected in various ion channels, most notably Shaker, and permitted the development of meticulous computational simulations of the activation mechanism. The fundamental premise for the structural interpretation of functional measurements from Shaker is that this channel and Kv1.2 have the same characteristics, such that correlation of data from both channels would be a trivial task. We tested these assumptions by measuring Kv1.2 voltage-dependent gating and charge per channel. We found that the Kv1.2 gating charge is near 10 elementary charges (eo), ∼25% less than the well-established 13-14 eo in Shaker. Next, we neutralized positive residues in the Kv1.2 S4 transmembrane segment to investigate the cause of the reduction of the gating charge and found that, whereas replacing R1 with glutamine decreased voltage sensitivity to ∼50% of the wild-type channel value, mutation of the subsequent arginines had a much smaller effect. These data are in marked contrast to the effects of charge neutralization in Shaker, where removal of the first four basic residues reduces the gating charge by roughly the same amount. In light of these differences, we propose that the voltage-sensing domains (VSDs) of Kv1.2 and Shaker might undergo the same physical movement, but the septum that separates the aqueous crevices in the VSD of Kv1.2 might be thicker than Shaker's, accounting for the smaller Kv1.2 gating charge.