Prospective validation of a transcriptomic severity classifier among patients with suspected acute infection and sepsis in the emergency department.

BACKGROUND AND IMPORTANCE: mRNA-based host response signatures have been reported to improve sepsis diagnostics. Meanwhile, prognostic markers for the rapid and accurate prediction of severity in patients with suspected acute infections and sepsis remain an unmet need. IMX-SEV-2 is a 29-host-mRNA classifier designed to predict disease severity in patients with acute infection or sepsis. OBJECTIVE: Validation of the host-mRNA infection severity classifier IMX-SEV-2. DESIGN, SETTINGS AND PARTICIPANTS: Prospective, observational, convenience cohort of emergency department (ED) patients with suspected acute infections. OUTCOME MEASURES AND ANALYSIS: Whole blood RNA tubes were analyzed using independently trained and validated composite target genes (IMX-SEV-2). IMX-SEV-2-generated risk scores for severity were compared to the patient outcomes in-hospital mortality and 72-h multiorgan failure. MAIN RESULTS: Of the 312 eligible patients, 22 (7.1%) died in hospital and 58 (18.6%) experienced multiorgan failure within 72 h of presentation. For predicting in-hospital mortality, IMX-SEV-2 had a significantly higher area under the receiver operating characteristic (AUROC) of 0.84 [95% confidence intervals (CI), 0.76-0.93] compared to 0.76 (0.64-0.87) for lactate, 0.68 (0.57-0.79) for quick Sequential Organ Failure Assessment (qSOFA) and 0.75 (0.65-0.85) for National Early Warning Score 2 (NEWS2), ( P = 0.015, 0.001 and 0.013, respectively). For identifying and predicting 72-h multiorgan failure, the AUROC of IMX-SEV-2 was 0.76 (0.68-0.83), not significantly different from lactate (0.73, 0.65-0.81), qSOFA (0.77, 0.70-0.83) or NEWS2 (0.81, 0.75-0.86). CONCLUSION: The IMX-SEV-2 classifier showed a superior prediction of in-hospital mortality compared to biomarkers and clinical scores among ED patients with suspected infections. No improvement for predicting multiorgan failure was found compared to established scores or biomarkers. Identifying patients with a high risk of mortality or multiorgan failure may improve patient outcomes, resource utilization and guide therapy decision-making.

A 6-mRNA host response classifier in whole blood predicts outcomes in COVID-19 and other acute viral infections.

Predicting the severity of COVID-19 remains an unmet medical need. Our objective was to develop a blood-based host-gene-expression classifier for the severity of viral infections and validate it in independent data, including COVID-19. We developed a logistic regression-based classifier for the severity of viral infections and validated it in multiple viral infection settings including COVID-19. We used training data (N = 705) from 21 retrospective transcriptomic clinical studies of influenza and other viral illnesses looking at a preselected panel of host immune response messenger RNAs. We selected 6 host RNAs and trained logistic regression classifier with a cross-validation area under curve of 0.90 for predicting 30-day mortality in viral illnesses. Next, in 1417 samples across 21 independent retrospective cohorts the locked 6-RNA classifier had an area under curve of 0.94 for discriminating patients with severe vs. non-severe infection. Next, in independent cohorts of prospectively (N = 97) and retrospectively (N = 100) enrolled patients with confirmed COVID-19, the classifier had an area under curve of 0.89 and 0.87, respectively, for identifying patients with severe respiratory failure or 30-day mortality. Finally, we developed a loop-mediated isothermal gene expression assay for the 6-messenger-RNA panel to facilitate implementation as a rapid assay. With further study, the classifier could assist in the risk assessment of COVID-19 and other acute viral infections patients to determine severity and level of care, thereby improving patient management and reducing healthcare burden.

A Novel 29-Messenger RNA Host-Response Assay From Whole Blood Accurately Identifies Bacterial and Viral Infections in Patients Presenting to the Emergency Department With Suspected Infections: A Prospective Observational Study.

OBJECTIVES: The rapid diagnosis of acute infections and sepsis remains a serious challenge. As a result of limitations in current diagnostics, guidelines recommend early antimicrobials for suspected sepsis patients to improve outcomes at a cost to antimicrobial stewardship. We aimed to develop and prospectively validate a new, 29-messenger RNA blood-based host-response classifier Inflammatix Bacterial Viral Non-Infected version 2 (IMX-BVN-2) to determine the likelihood of bacterial and viral infections. DESIGN: Prospective observational study. SETTING: Emergency Department, Campus Benjamin Franklin, Charité-Universitätsmedizin Berlin, Germany. PATIENTS: Three hundred twelve adult patients presenting to the emergency department with suspected acute infections or sepsis with at least one vital sign change. INTERVENTIONS: None (observational study only). MEASUREMENTS AND MAIN RESULTS: Gene expression levels from extracted whole blood RNA was quantified on a NanoString nCounter SPRINT (NanoString Technologies, Seattle, WA). Two predicted probability scores for the presence of bacterial and viral infection were calculated using the IMX-BVN-2 neural network classifier, which was trained on an independent development set. The IMX-BVN-2 bacterial score showed an area under the receiver operating curve for adjudicated bacterial versus ruled out bacterial infection of 0.90 (95% CI, 0.85-0.95) compared with 0.89 (95% CI, 0.84-0.94) for procalcitonin with procalcitonin being used in the adjudication. The IMX-BVN-2 viral score area under the receiver operating curve for adjudicated versus ruled out viral infection was 0.83 (95% CI, 0.77-0.89). CONCLUSIONS: IMX-BVN-2 demonstrated accuracy for detecting both viral infections and bacterial infections. This shows the potential of host-response tests as a novel and practical approach for determining the causes of infections, which could improve patient outcomes while upholding antimicrobial stewardship.

Optimization of Genomic Classifiers for Clinical Deployment: Evaluation of Bayesian Optimization to Select Predictive Models of Acute Infection and In-Hospital Mortality.

Acute infection, if not rapidly and accurately detected, can lead to sepsis, organ failure and even death. Current detection of acute infection as well as assessment of a patient's severity of illness are imperfect. Characterization of a patient's immune response by quantifying expression levels of specific genes from blood represents a potentially more timely and precise means of accomplishing both tasks. Machine learning methods provide a platform to leverage this host response for development of deployment-ready classification models. Prioritization of promising classifiers is dependent, in part, on hyperparameter optimization for which a number of approaches including grid search, random sampling and Bayesian optimization have been shown to be effective. We compare HO approaches for the development of diagnostic classifiers of acute infection and in-hospital mortality from gene expression of 29 diagnostic markers. We take a deployment-centered approach to our comprehensive analysis, accounting for heterogeneity in our multi-study patient cohort with our choices of dataset partitioning and hyperparameter optimization objective as well as assessing selected classifiers in external (as well as internal) validation. We find that classifiers selected by Bayesian optimization for in-hospital mortality can outperform those selected by grid search or random sampling. However, in contrast to previous research: 1) Bayesian optimization is not more efficient in selecting classifiers in all instances compared to grid search or random sampling-based methods and 2) we note marginal gains in classifier performance in only specific circumstances when using a common variant of Bayesian optimization (i.e. automatic relevance determination). Our analysis highlights the need for further practical, deployment-centered benchmarking of HO approaches in the healthcare context.

A generalizable 29-mRNA neural-network classifier for acute bacterial and viral infections.

Improved identification of bacterial and viral infections would reduce morbidity from sepsis, reduce antibiotic overuse, and lower healthcare costs. Here, we develop a generalizable host-gene-expression-based classifier for acute bacterial and viral infections. We use training data (N = 1069) from 18 retrospective transcriptomic studies. Using only 29 preselected host mRNAs, we train a neural-network classifier with a bacterial-vs-other area under the receiver-operating characteristic curve (AUROC) 0.92 (95% CI 0.90-0.93) and a viral-vs-other AUROC 0.92 (95% CI 0.90-0.93). We then apply this classifier, inflammatix-bacterial-viral-noninfected-version 1 (IMX-BVN-1), without retraining, to an independent cohort (N = 163). In this cohort, IMX-BVN-1 AUROCs are: bacterial-vs.-other 0.86 (95% CI 0.77-0.93), and viral-vs.-other 0.85 (95% CI 0.76-0.93). In patients enrolled within 36 h of hospital admission (N = 70), IMX-BVN-1 AUROCs are: bacterial-vs.-other 0.92 (95% CI 0.83-0.99), and viral-vs.-other 0.91 (95% CI 0.82-0.98). With further study, IMX-BVN-1 could provide a tool for assessing patients with suspected infection and sepsis at hospital admission.

Modulation of Macropinocytosis-Mediated Internalization Decreases Ocular Toxicity of Antibody-Drug Conjugates.

AGS-16C3F is an antibody-drug conjugate (ADC) against ectonucleotide pyrophosphatase/phosphodiesterase 3 (ENPP3) containing the mcMMAF linker-payload currently in development for treatment of metastatic renal cell carcinoma. AGS-16C3F and other ADCs have been reported to cause ocular toxicity in patients by unknown mechanisms. To investigate this toxicity, we developed an in vitro assay using human corneal epithelial cells (HCEC) and show that HCECs internalized AGS-16C3F and other ADCs by macropinocytosis, causing inhibition of cell proliferation. We observed the same mechanism for target-independent internalization of AGS-16C3F in fibroblasts and human umbilical vein endothelial cells (HUVEC). Macropinocytosis-mediated intake of macromolecules is facilitated by the presence of positive charges or hydrophobic residues on the surface of the macromolecule. Modification of AGS-16C3F, either by attachment of poly-glutamate peptides, mutation of residue K16 to D on AGS-16C3F [AGS-16C3F(K16D)], or decreasing the overall hydrophobicity via attachment of polyethylene glycol moieties, significantly reduced cytotoxicity against HCECs and other primary cells. Rabbits treated with AGS-16C3F showed significant ocular toxicity, whereas those treated with AGS-16C3F(K16D) presented with less severe and delayed toxicities. Both molecules displayed similar antitumor activity in a mouse xenograft model. These findings establish a mechanism of action for target-independent toxicities of AGS-16C3F and ADCs in general, and provide methods to ameliorate these toxicities.Significance: These findings reveal a mechanism for nonreceptor-mediated toxicities of antibody drug conjugates and potential solutions to alleviate these toxicities. Cancer Res; 78(8); 2115-26. ©2018 AACR.

Quantitative proteomic profiling identifies protein correlates to EGFR kinase inhibition.

Clinical oncology is hampered by lack of tools to accurately assess a patient's response to pathway-targeted therapies. Serum and tumor cell surface proteins whose abundance, or change in abundance in response to therapy, differentiates patients responding to a therapy from patients not responding to a therapy could be usefully incorporated into tools for monitoring response. Here, we posit and then verify that proteomic discovery in in vitro tissue culture models can identify proteins with concordant in vivo behavior and further, can be a valuable approach for identifying tumor-derived serum proteins. In this study, we use stable isotope labeling of amino acids in culture (SILAC) with proteomic technologies to quantitatively analyze the gefitinib-related protein changes in a model system for sensitivity to EGF receptor (EGFR)-targeted tyrosine kinase inhibitors. We identified 3,707 intracellular proteins, 1,276 cell surface proteins, and 879 shed proteins. More than 75% of the proteins identified had quantitative information, and a subset consisting of 400 proteins showed a statistically significant change in abundance following gefitinib treatment. We validated the change in expression profile in vitro and screened our panel of response markers in an in vivo isogenic resistant model and showed that these were markers of gefitinib response and not simply markers of phospho-EGFR downregulation. In doing so, we also were able to identify which proteins might be useful as markers for monitoring response and which proteins might be useful as markers for a priori prediction of response.

Precursor-ion mass re-estimation improves peptide identification on hybrid instruments.

Mass spectrometry-based proteomics experiments have become an important tool for studying biological systems. Identifying the proteins in complex mixtures by assigning peptide fragmentation spectra to peptide sequences is an important step in the proteomics process. The 1-2 ppm mass-accuracy of hybrid instruments, like the LTQ-FT, has been cited as a key factor in their ability to identify a larger number of peptides with greater confidence than competing instruments. However, in replicate experiments of an 18-protein mixture, we note parent masses deviate 171 ppm, on average, for ion-trap data directed identifications and 8 ppm, on average, for preview Fourier transform (FT) data directed identifications. These deviations are neither caused by poor calibration nor by excessive ion-loading and are most likely due to errors in parent mass estimation. To improve these deviations, we introduce msPrefix, a program to re-estimate a peptide's parent mass from an associated high-accuracy full-scan survey spectrum. In 18-protein mixture experiments, msPrefix parent mass estimates deviate only 1 ppm, on average, from the identified peptides. In a cell lysate experiment searched with a tolerance of 50 ppm, 2295 peptides were confidently identified using native data and 4560 using msPrefixed data. Likewise, in a plasma experiment searched with a tolerance of 50 ppm, 326 peptides were identified using native data and 1216 using msPrefixed data. msPrefix is also able to determine which MS/MS spectra were possibly derived from multiple precursor ions. In complex mixture experiments, we demonstrate that more than 50% of triggered MS/MS may have had multiple precursor ions and note that spectra with multiple candidate ions are less likely to result in an identification using TANDEM. These results demonstrate integration of msPrefix into traditional shotgun proteomics workflows significantly improves identification results.