Machine Learning in Laboratory Medicine: Recommendations of the IFCC Working Group.

BACKGROUND: Machine learning (ML) has been applied to an increasing number of predictive problems in laboratory medicine, and published work to date suggests that it has tremendous potential for clinical applications. However, a number of groups have noted the potential pitfalls associated with this work, particularly if certain details of the development and validation pipelines are not carefully controlled. METHODS: To address these pitfalls and other specific challenges when applying machine learning in a laboratory medicine setting, a working group of the International Federation for Clinical Chemistry and Laboratory Medicine was convened to provide a guidance document for this domain. RESULTS: This manuscript represents consensus recommendations for best practices from that committee, with the goal of improving the quality of developed and published ML models designed for use in clinical laboratories. CONCLUSIONS: The committee believes that implementation of these best practices will improve the quality and reproducibility of machine learning utilized in laboratory medicine. SUMMARY: We have provided our consensus assessment of a number of important practices that are required to ensure that valid, reproducible machine learning (ML) models can be applied to address operational and diagnostic questions in the clinical laboratory. These practices span all phases of model development, from problem formulation through predictive implementation. Although it is not possible to exhaustively discuss every potential pitfall in ML workflows, we believe that our current guidelines capture best practices for avoiding the most common and potentially dangerous errors in this important emerging field.

Methodologies in Mitochondrial Testing: Diagnosing a Primary Mitochondrial Respiratory Chain Disorder.

BACKGROUND: Mitochondria are cytosolic organelles within most eukaryotic cells. Mitochondria generate the majority of cellular energy in the form of adenosine triphosphate (ATP) through oxidative phosphorylation (OxPhos). Pathogenic variants in mitochondrial DNA (mtDNA) and nuclear DNA (nDNA) lead to defects in OxPhos and physiological malfunctions (Nat Rev Dis Primer 2016;2:16080.). Patients with primary mitochondrial disorders (PMD) experience heterogeneous symptoms, typically in multiple organ systems, depending on the tissues affected by mitochondrial dysfunction. Because of this heterogeneity, clinical diagnosis is challenging (Annu Rev Genomics Hum Genet 2017;18:257-75.). Laboratory diagnosis of mitochondrial disease depends on a multipronged analysis that can include biochemical, histopathologic, and genetic testing. Each of these modalities has complementary strengths and limitations in diagnostic utility. CONTENT: The primary focus of this review is on diagnosis and testing strategies for primary mitochondrial diseases. We review tissue samples utilized for testing, metabolic signatures, histologic findings, and molecular testing approaches. We conclude with future perspectives on mitochondrial testing. SUMMARY: This review offers an overview of the current biochemical, histologic, and genetic approaches available for mitochondrial testing. For each we review their diagnostic utility including complementary strengths and weaknesses. We identify gaps in current testing and possible future avenues for test development.

2-Methylglutaconic acid as a biomarker in routine urine organic acids leading to the diagnosis of glutaric acidemia type I in a low excretor.

GA1 (OMIM# 231670) is an organic aciduria caused by defective Glutaryl-CoA dehydrogenase (GCDH), encoded by GCDH. Early detection of GA1 is crucial to prevent patients from developing acute encephalopathic crisis and subsequent neurologic sequelae. Diagnosis of GA1 relies on elevated glutarylcarnitine (C5DC) in plasma acylcarnitine analysis and hyperexcretion of glutaric acid (GA) and 3-hydroxyglutaric acid (3HG) in urine organic acid (UOA) analysis. Low excretors (LE), however, exhibit subtly elevated or even normal plasma C5DC and urinary GA levels, leading to screening and diagnostic challenges. The measurement of 3HG in UOA is thus often used as the 1st tier test for GA1. We described a case of LE detected via newborn screen with normal excretion of GA, absent of 3HG and increased 2-methylglutaconic acid (2MGA), which was detected at 3 mg/g creatinine (reference interval <1 mg/g creatinine) without appreciable ketones. We retrospectively examined UOA of 8 other GA1 patients and the 2MGA level ranged from 2.5 to 27.39 mg/g creatinine, which is significantly higher than normal controls (0.05-1.61 mg/g creatinine). Although the underlying mechanism of 2MGA formation in GA1 is unclear, our study suggests 2MGA is a biomarker for GA1 and should be monitored by routine UOA to evaluate its diagnostic and prognostic value.

Rap1 relocalization contributes to the chromatin-mediated gene expression profile and pace of cell senescence.

Cellular senescence is accompanied by dramatic changes in chromatin structure and gene expression. Using Saccharomyces cerevisiae mutants lacking telomerase (tlc1Δ) to model senescence, we found that with critical telomere shortening, the telomere-binding protein Rap1 (repressor activator protein 1) relocalizes to the upstream promoter regions of hundreds of new target genes. The set of new Rap1 targets at senescence (NRTS) is preferentially activated at senescence, and experimental manipulations of Rap1 levels indicate that it contributes directly to NRTS activation. A notable subset of NRTS includes the core histone-encoding genes; we found that Rap1 contributes to their repression and that histone protein levels decline at senescence. Rap1 and histones also display a target site-specific antagonism that leads to diminished nucleosome occupancy at the promoters of up-regulated NRTS. This antagonism apparently impacts the rate of senescence because underexpression of Rap1 or overexpression of the core histones delays senescence. Rap1 relocalization is not a simple consequence of lost telomere-binding sites, but rather depends on the Mec1 checkpoint kinase. Rap1 relocalization is thus a novel mechanism connecting DNA damage responses (DDRs) at telomeres to global changes in chromatin and gene expression while driving the pace of senescence.

Derivation of a metabolic signature associated with bacterial meningitis in infants.

BACKGROUND: Diagnosis of bacterial meningitis (BM) is challenging in newborn infants. Presently, biomarkers of BM have limited diagnostic accuracy. Analysis of cerebrospinal fluid (CSF) metabolites may be a useful diagnostic tool in BM. METHODS: In a nested case-control study, we examined >400 metabolites in CSF of uninfected infants and infants with culture-confirmed BM using gas and liquid chromatography mass spectrometry. Preterm and full-term infants in a Level III or IV Neonatal Intensive Care Unit were prospectively enrolled when evaluated for serious bacterial infection. RESULTS: Over 200 CSF metabolites significantly differed in uninfected infants and infants with BM. Using machine learning, we found that as few as 6 metabolites distinguished infants with BM from uninfected infants in this pilot cohort. Further analysis demonstrated three metabolites associated with Group B Streptococcal meningitis. CONCLUSIONS: We report the first comprehensive metabolic analysis of CSF in infants with BM. In our pilot cohort, we derived a metabolic signature that predicted the presence or absence of BM, irrespective of gestational age, postnatal age, sex, race and ethnicity, presence of neurosurgical hardware, white blood cell count in CSF, and red blood cell contamination in CSF. Metabolic analysis may aid diagnosis of BM and facilitate clinical decision-making in infants. IMPACT: In a pilot cohort, metabolites in cerebrospinal fluid distinguished infants with bacterial meningitis from uninfected infants.We report the first comprehensive metabolic analysis of cerebrospinal fluid in infants with bacterial meningitis.Our findings may be used to improve diagnosis of bacterial meningitis and to offer mechanistic insights into the pathophysiology of bacterial meningitis in infants.

Vision and Vestibular System Dysfunction Predicts Prolonged Concussion Recovery in Children.

OBJECTIVE: Up to one-third of children with concussion have prolonged symptoms lasting beyond 4 weeks. Vision and vestibular dysfunction is common after concussion. It is unknown whether such dysfunction predicts prolonged recovery. We sought to determine which vision or vestibular problems predict prolonged recovery in children. DESIGN: A retrospective cohort of pediatric patients with concussion. SETTING: A subspecialty pediatric concussion program. PATIENTS (OR PARTICIPANTS): Four hundred thirty-two patient records were abstracted. ASSESSMENT OF RISK FACTORS: Presence of vision or vestibular dysfunction upon presentation to the subspecialty concussion program. MAIN OUTCOME MEASURES: The main outcome of interest was time to clinical recovery, defined by discharge from clinical follow-up, including resolution of acute symptoms, resumption of normal physical and cognitive activity, and normalization of physical examination findings to functional levels. RESULTS: Study subjects were 5 to 18 years (median = 14). A total of 378 of 432 subjects (88%) presented with vision or vestibular problems. A history of motion sickness was associated with vestibular dysfunction. Younger age, public insurance, and presence of headache were associated with later presentation for subspecialty concussion care. Vision and vestibular problems were associated within distinct clusters. Provocable symptoms with vestibulo-ocular reflex (VOR) and smooth pursuits and abnormal balance and accommodative amplitude (AA) predicted prolonged recovery time. CONCLUSIONS: Vision and vestibular problems predict prolonged concussion recovery in children. A history of motion sickness may be an important premorbid factor. Public insurance status may represent problems with disparities in access to concussion care. Vision assessments in concussion must include smooth pursuits, saccades, near point of convergence (NPC), and accommodative amplitude (AA). A comprehensive, multidomain assessment is essential to predict prolonged recovery time and enable active intervention with specific school accommodations and targeted rehabilitation.

Near Point of Convergence after Concussion in Children.

PURPOSE: To determine the incidence of abnormal near point of convergence (NPC) after acute concussion in pediatric patients and to describe the clinical course of such patients. METHODS: A retrospective cohort study of 275 pediatric patients 5 to 18 years of age presenting to a tertiary care children's hospital subspecialty concussion program with a new concussion between July 1, 2014 and June 30, 2015 was conducted. RESULTS: Sixty-seven out of 275 pediatric patients presenting to a subspecialty concussion referral program were found to have abnormal NPC on physical examination as measured by an accommodative rule. Twenty-six (46%) patients recovered with standard clinical care over a median time frame of 4.5 weeks (range 1-18), including a brief period of cognitive and physical rest followed by gradual return to school and physical activities without any formal interventions. An additional 23 (41%) patients recovered a median of 11 weeks post-injury after referral for formal vestibular therapy, including interventions for abnormal convergence, such as Brock string and pencil pushups. Seven (13%) patients with persistent abnormal NPC and concomitant symptoms that necessitated referral for formal office-based vision therapy with developmental optometry recovered a median of 23 weeks post-injury and a median of 16 weeks after referral to vision therapy. CONCLUSIONS: Assessment for NPC is a diagnostic entity that warrants consideration in children with concussion. Concussion questionnaires may not be sensitive to detect vision symptoms in children, making an accurate assessment for convergence important in the evaluation of concussion. Some children with abnormal NPC will recover without any formal intervention after concussion; however, a subset of patients with persistent abnormal NPC after concussion may benefit from interventions including vestibular and/or vision therapy.

Identification of prolyl hydroxylation modifications in mammalian cell proteins.

Prolyl hydroxylation is a PTM that plays an important role in the formation of collagen fibrils and in the oxygen-dependent regulation of hypoxia inducible factor-α (HIF-α). While this modification has been well characterized in the context of these proteins, it remains unclear to what extent it occurs in the remaining mammalian proteome. We explored this question using MS to analyze cellular extracts subjected to various fractionation strategies. In one strategy, we employed the von Hippel Lindau tumor suppressor protein, which recognizes prolyl hydroxylated HIF-α, as a scaffold for generating hydroxyproline capture reagents. We report novel sites of prolyl hydroxylation within five proteins: FK506-binding protein 10, myosin heavy chain 10, hexokinase 2, pyruvate kinase, and C-1 Tetrahydrofolate synthase. Furthermore, we show that identification of prolyl hydroxylation presents a significant technical challenge owing to widespread isobaric methionine oxidation, and that manual inspection of spectra of modified peptides in this context is critical for validation.

Defective Tibetan PHD2 binding to p23 links high altitude adaption to altered oxygen sensing.

The Tibetan population has adapted to the chronic hypoxia of high altitude. Tibetans bear a genetic signature in the prolyl hydroxylase domain protein 2 (PHD2/EGLN1) gene, which encodes for the central oxygen sensor of the hypoxia-inducible factor (HIF) pathway. Recent studies have focused attention on two nonsynonymous coding region substitutions, D4E and C127S, both of which are markedly enriched in the Tibetan population. These amino acids reside in a region of PHD2 that harbors a zinc finger, which we have previously discovered binds to a Pro-Xaa-Leu-Glu (PXLE) motif in the HSP90 cochaperone p23, thereby recruiting PHD2 to the HSP90 pathway to facilitate HIF-α hydroxylation. We herein report that the Tibetan PHD2 haplotype (D4E/C127S) strikingly diminishes the interaction of PHD2 with p23, resulting in impaired PHD2 down-regulation of the HIF pathway. The defective binding to p23 depends on both the D4E and C127S substitutions. We also identify a PXLE motif in HSP90 itself that can mediate binding to PHD2 but find that this interaction is maintained with the D4E/C127S PHD2 haplotype. We propose that the Tibetan PHD2 variant is a loss of function (hypomorphic) allele, leading to augmented HIF activation to facilitate adaptation to high altitude.

Validation of DNA methylation to predict outcome in acute myeloid leukemia by use of xMELP.

BACKGROUND: Epigenetic dysregulation involving alterations in DNA methylation is a hallmark of various types of cancer, including acute myeloid leukemia (AML). Although specific cancer types and clinical aggressiveness of tumors can be determined by DNA methylation status, the assessment of DNA methylation at multiple loci is not routinely performed in the clinical laboratory. METHODS: We recently described a novel microsphere-based assay for multiplex evaluation of DNA methylation. In the current study, we validated and used an improved assay [termed expedited microsphere HpaII small fragment Enrichment by Ligation-mediated PCR (xMELP)] that can be performed with appropriate clinical turnaround time. RESULTS: Using the xMELP assay in conjunction with a new 17-locus random forest classifier that has been trained using 344 AML samples, we were able to segregate an independent cohort of 70 primary AML patients into methylation-determined subgroups with significantly distinct mortality risk (P = 0.009). We also evaluated precision, QC parameters, and preanalytic variables of the xMELP assay and determined the sensitivity of the random forest classifier score to failure at 1 or more loci. CONCLUSIONS: Our results demonstrate that xMELP performance is suitable for implementation in the clinical laboratory and predicts AML outcome in an independent patient cohort.

Urine neutrophil gelatinase-associated lipocalin and risk of cardiovascular disease and death in CKD: results from the Chronic Renal Insufficiency Cohort (CRIC) Study.

BACKGROUND: Chronic kidney disease is common and is associated with increased cardiovascular disease risk. Currently, markers of renal tubular injury are not used routinely to describe kidney health and little is known about the risk of cardiovascular events and death associated with these biomarkers independent of glomerular filtration-based markers (such as serum creatinine or albuminuria). STUDY DESIGN: Cohort study, CRIC (Chronic Renal Insufficiency Cohort) Study. SETTING & PARTICIPANTS: 3,386 participants with estimated glomerular filtration rate of 20 to 70mL/min/1.73m(2) enrolled from June 2003 through August 2008. PREDICTOR: Urine neutrophil gelatinase-associated lipocalin (NGAL) concentration. OUTCOMES: Adjudicated heart failure event, ischemic atherosclerotic event (myocardial infarction, ischemic stroke, or peripheral artery disease), and death through March 2011. MEASUREMENTS: Urine NGAL measured at baseline with a 2-step assay using chemiluminescent microparticle immunoassay technology on an ARCHITECT i2000SR (Abbott Laboratories). RESULTS: There were 428 heart failure events (during 16,383 person-years of follow-up), 361 ischemic atherosclerotic events (during 16,584 person-years of follow-up), and 522 deaths (during 18,214 person-years of follow-up). In Cox regression models adjusted for estimated glomerular filtration rate, albuminuria, demographics, traditional cardiovascular disease risk factors, and cardiac medications, higher urine NGAL levels remained associated independently with ischemic atherosclerotic events (adjusted HR for the highest [>49.5ng/mL] vs lowest [≤6.9ng/mL] quintile, 1.83 [95% CI, 1.20-2.81]; HR per 0.1-unit increase in log urine NGAL, 1.012 [95% CI, 1.001-1.023]), but not heart failure events or deaths. LIMITATIONS: Urine NGAL was measured only once. CONCLUSIONS: Among patients with chronic kidney disease, urine levels of NGAL, a marker of renal tubular injury, were associated independently with future ischemic atherosclerotic events, but not with heart failure events or deaths.

Serum Fractalkine (CX3CL1) and Cardiovascular Outcomes and Diabetes: Findings From the Chronic Renal Insufficiency Cohort (CRIC) Study.

BACKGROUND: Cardiometabolic disease is a major cause of morbidity and mortality in persons with chronic kidney disease (CKD). Fractalkine (CX3CL1) is a potential mediator of both atherosclerosis and metabolic disease. Studies of the relationship of CX3CL1 with risk of cardiovascular disease (CVD) events and metabolic traits are lacking, particularly in the high-risk setting of CKD. STUDY DESIGN: Cross-sectional and longitudinal observational analysis. SETTING & PARTICIPANTS: Adults with CKD from 7 US sites participating in the Chronic Renal Insufficiency Cohort (CRIC) Study. PREDICTOR: Quartiles of plasma CX3CL1 levels at baseline. OUTCOMES: Baseline estimated glomerular filtration rate from a creatinine and cystatin C-based equation, prevalent and incident CVD, diabetes, metabolic syndrome and its criteria, homeostatic model assessment of insulin resistance, hemoglobin A1c level, myocardial infarction, all-cause mortality, and the composite outcome of myocardial infarction/all-cause mortality. RESULTS: Among 3,687 participants, baseline CX3CL1 levels were associated positively with several CVD risk factors and metabolic traits, lower estimated glomerular filtration rate, and higher levels of inflammatory cytokines, as well as prevalent CVD (OR, 1.09; 95% CI, 1.01-1.19; P=0.03). Higher CX3CL1 level also was associated with prevalent diabetes (OR, 1.26; 95% CI, 1.16-1.38; P<0.001) in adjusted models. During a mean follow-up of 6 years, there were 352 deaths, 176 myocardial infarctions, and 484 composite outcomes. In fully adjusted models, 1-SD higher CX3CL1 level increased the hazard for all-cause mortality (1.11; 95% CI, 1.00-1.22; P=0.02) and the composite outcome (1.09; 95% CI, 1.00-1.19; P=0.04). LIMITATIONS: Study design did not allow evaluation of changes over time, correlation with progression of phenotypes, or determination of causality of effect. CONCLUSIONS: Circulating CX3CL1 level may contribute to both atherosclerotic CVD and diabetes in a CKD cohort. Further studies are required to establish mechanisms through which CX3CL1 affects the pathogenesis of atherosclerosis and diabetes.