key: cord-1040275-2gxbm3p2
authors: Sindelar, Miriam; Stancliffe, Ethan; Schwaiger-Haber, Michaela; Anbukumar, Dhanalakshmi S.; Adkins-Travis, Kayla; Goss, Charles W.; O'Halloran, Jane A.; Mudd, Philip A.; Liu, Wen-Chun; Albrecht, Randy A.; García-Sastre, Adolfo; Shriver, Leah P.; Patti, Gary J.
title: Longitudinal Metabolomics of Human Plasma Reveals Prognostic Markers of COVID-19 Disease Severity
date: 2021-07-21
journal: Cell Rep Med
DOI: 10.1016/j.xcrm.2021.100369
sha: 9aee791c49851c9f4d2baaea1283265714d802c9
doc_id: 1040275
cord_uid: 2gxbm3p2

There is an urgent need to identify which COVID-19 patients will develop life-threatening illness so that medical resources can be optimally allocated and rapid treatment can be administered early in the disease course, when clinical management is most effective. To aid in the prognostic classification of disease severity, we perform untargeted metabolomics on plasma from 339 patients, with samples collected at six longitudinal time points. Using the temporal metabolic profiles and machine learning, we build a predictive model of disease severity. We discover that a panel of metabolites measured at the time of study entry successfully determine disease severity. Through analysis of longitudinal samples, we confirm that the majority of these markers are directly related to disease progression and that their levels return to baseline upon disease recovery. Finally, we validate that these metabolites are also altered in a hamster model of COVID-19.

Coronavirus disease 2019 , which is caused by infection with the novel 51 coronavirus SARS-CoV-2, has led to a global health crisis 1 . To date, more than 150 million cases 52 of COVID-19 have been reported worldwide and resulted in over 3.2 million deaths 2 . The 53 infection-fatality rate of SARS-CoV-2 can be reduced with the appropriate care (e.g., intensive 54 care unit beds, oxygen, staff, extracorporeal-life support, and therapeutics). Unfortunately, 55 hospital resources can quickly become depleted in situations where cases spike 3 . Although 56 vaccination efforts are underway worldwide, SARS-CoV-2 infections continue to increase 57 rapidly in a number of countries such as India and Brazil, where medical facilities are being 58 overwhelmed 4 . 59 Patients who develop critical illness from COVID-19 are best treated early in the disease 60 course before the onset of severe symptoms 5, 6 . It is currently difficult to determine which subset 61 of patients will develop life-threatening disease and therefore most benefit from receiving 62 treatment when resources are limited 7 . Early identification of these patients would allow optimal 63 allocation of care. To this end, the objective of the current study was to identify metabolites in 64 patient plasma that accurately predict life-threatening cases of COVID-19 prior to the onset of Out of 272 COV+ patients, 242 showed at least one COVID-19-related symptom mentioned 111 by the Centers for Disease Control and Prevention (CDC) including fever, chills, conjunctival 112 congestion, nasal congestion, headaches, cough, sore throat, shortness of breath, nausea or 113 vomiting, diarrhea, myalgia, fatigue, and loss of taste or smell 27 . Of the remaining COV+ cases, 114 21 showed no symptoms and were classified as asymptomatic. Nine subjects in the COV+ group 115 displayed other symptoms such as confusion, lethargy, an altered mental state, or breathing 116 anomalies. Out of the 67 COV-cases, 65 presented with at least one COVID-19-related 117 symptom upon study entry, while two received a SARS-CoV-2-PCR test upon exposure to a 118 SARS-CoV-2-positive individual. The frequency of COVID-19-related symptoms is shown in 119 Table 1, and the distributions across the COV+ and COV-cohorts are depicted in Figure S1C . In 120 both the COV-and COV+ groups, the number of COVID-19 related symptoms reported per 121 individual was comparable. The breakdown of how many symptoms were experienced per 122 individual in both the COV+ and COV-groups is shown in Figure S1D -E. 123 Next, we examined the distribution of comorbidities and self-reported medical history 124 presented in the WU350 cohort. Medical conditions were assigned based on ICD10 codes and 125 the Elixhauser naming conventions 28 . Chronic kidney disease (CKD) and diabetes (recorded up 126 to one year prior to the current admission and up to one year prior to the d0 sample for those who 127 were not hospitalized) was significantly higher in the COV+ group compared to the COV-group 128 ( Table 1 ). The incidence of acute renal failure and acute respiratory failure during the present 129 hospitalization was also significantly elevated in the COV+ group compared to the COV-group 1 Includes both training and test cohort, percentages are shown as % of the group (COV-or COV+)Demographic data last updated May 20 th 2021. 2 Data are presented as mean ± standard deviation, p-values of numeric parameters were calculated using a 2-tailed Student's t-test with unequal variance, p-values of categorical parameters were calculated using a chi-square test. 3 Abbreviations: Mmale, Ffemale, yryears, N/Anot applicable 4 Two SARS-CoV-2 negative individuals were without symptoms but exposed to a SARS-CoV-2-positive individual. Nine SARS-CoV-2 positive individuals had other symptoms (e.g., confusion, lethargy, an altered mental state, or breathing anomalies). Clinical metadata last updated October 16 th 2020. 5 CDC guideline symptom was added to the symptom questionnaire late in the study, parameter is not available for the majority of the subjects 6 A SARS-CoV-2 test was routinely administered at presentation to the hospital for reasons other than the COVID-19 disease (e.g., accidents, pre-operation tests, regular checkups, cancer screening, injuries, or exposure to a SARS-CoV-2-positive individual). 7 During present hospitalization 8 Recorded up to one year prior to the current admission or up to one year prior to the d0 sample for those who were not admitted to the hospital 9 At any point during hospital stay 10 Hospital and/or ICU admission of COV-group for reasons other than COVID-19 (e.g., accidents, acute respiratory failure due to bacterial pneumonia, intentional self-harm, possible heart failure, hypertension, trauma, cancer). 11 Percentages shown as % of total deceased due to J o u r n a l P r e -p r o o f Figure S2C but no 165 longer clustered based on batch ( Figure S2D ). 166 The goal of this study was to find metabolic alterations that are predictive of disease severity 167 in SARS-CoV-2-positive individuals. We used admission to the ICU during disease progression 168 to classify patients as having severe or non-severe disease, as has been done previously 30, 31 dimethylguanosine 33 were also significantly altered ( Figure 1B and Figure S3 ). when the d0 sample was collected ( Figure S4A ). Although this diversity added additional 196 variance to the d0 metabolic signatures, it also enabled a more clinically relevant model to be 197 constructed as patients will present to the hospital at various points in the disease course. We 198 note that COV+ severe and non-severe patients were not significantly different in time from 199 symptom onset to the d0 sample collection ( Figure S4B ). The 707 metabolites that composed the 200 metabolic profiles served as the predictors for our ML model. To assess predictive power, we 201 split our dataset into two distinct groups of COV+ patients: a training set (163 patients) that we 202 used to select, optimize, and train our ML model and an independent test set (100 patients) that 203 was only used to evaluate the model's performance ( Figure 1C , Table S1 , Table S2 ). Using our 204 training set, we evaluated the efficacy of five ML algorithms with 20-fold cross validation and 205 found that a linear ElasticNet 34 regression model was the most effective ( Figure S4C ). After 206 training the model, we applied it to the patients in the test set and assessed performance by using 207 the area under the receiver operating characteristic curve (AUC). On the test set, we see strong 208 J o u r n a l P r e -p r o o f predictive performance (AUC = 0.72) that outperforms a simple model that only uses BMI and 209 age to predict disease severity ( Figure 2A ) and is significantly more predictive than a random 210 model ( Figure 2B , see Permutation test in Methods). As further validation, when the trained 211 model was applied to the COV-patients (no COV-patients were in the training set), the mean 212 scores output by the model were lower than those for the COV+ non-severe and the COV+ 213 severe patients in the test set ( Figure S4D ). This indicates that the model can not only 214 differentiate disease severity but also can distinguish COV+ and COV-patients. We wish to 215 emphasize that PCR is the gold standard to diagnose SARS-CoV-2 infection. As such, we 216 present this result only as confirmation that our model correctly predicts disease severity and not 217 as a potential diagnostic for viral infection. 218 We next sought to interpret which metabolites were most salient to the model's predictions. 219 First, we computed the variable importance of the model when trained on the complete dataset, 220 which found 92 unique metabolites that contributed to the model's predictions. Among this 221 group of 92 compounds were metabolites that have been previously implicated in SARS-CoV-2 222 infection such as kynurenate, nicotinamide, creatinine, LPCs, PCs, and others 11, 12, 14, 32 . The mean 223 intensity of each metabolite in the COV-, COV+ non-severe, and COV+ severe groups can be 224 seen in Figure S5 . Next, we aimed to assess the robustness of the metabolites selected by the ML better than a random model or a model that used only BMI and age as predictors ( Figure S6A -B). 234 To ensure that our model is indeed capable of predicting future disease severity, we additionally 235 evaluated our model on the portion of the test-set patients who were not admitted to the ICU on 236 or before when the d0 sample was collected. We again found that our model's performance 237 remains high (AUC = 0.71) and continues to outperform a model based on BMI and age ( Figure   238 S6C-D). The variable importance of the predictor metabolites when trained on the entire dataset 239 is shown in Figure 2C . The mean intensity of the metabolites in the COV-, COV+ non-severe, 240 and COV+ severe groups is shown in Figure 2D and Figure S7 . Additionally, the intensities of 241 these metabolites among test-set patients and patients not admitted to ICU on d0 is provided in 242 Figure S8 . To exclude the possibility that the metabolite alterations were due to the application 243 of therapeutics, we additionally removed all d0 samples that were administered dexamethasone 244 or remdesivir on the same day as sample collection. Using this subset, all predictor metabolites 245 were still significantly altered. All LPCs and PCs that contributed to the model, as well as serine, 246 presented a significant downward trend of signal abundance with disease severity. Conversely, 247 the other polar metabolites, (kynurenate and 1-methyladenosine), phosphatidylethanolamines 248 (PEs), and ceramides exhibited a significant upward trend in signal intensity ( Figure 2E ). (Table S3 ). The COV+ severe group is significantly older than the non-severe group ( Figure 3A ), 258 but there was no significant difference in BMI ( Figure 3B ). CO 2 levels were not significantly 259 altered between groups ( Figure 3C) , with values mostly being in the normal range. In contrast, 260 there were significantly increased levels of the inflammatory marker C-reactive protein (CRP, 261 Figure 3D ). D-dimer, absolute neutrophil count, and neutrophil % were also increased ( Figure   262 3E-G). Absolute lymphocyte count and lymphocyte % were decreased ( Figure 3H -I). These data 263 indicate more severe inflammation in the COV+ severe group compared to the non-severe group 264 and are consistent with reports from previous studies 11, 36, 37 . Neutrophil recruitment has also been 265 shown to be dysregulated in severe cases of COVID-19 38-41 . Lymphopenia, the abnormally low 266 levels of blood lymphocytes, has been found to correlate with disease severity of COVID-19 and 267 even to be predictive of disease severity 42-46 . 268 Given that specific comorbidities increase the risk of having a severe case of COVID-19 269 31,47,48 , we asked which co-morbidities are enriched in the COV+ severe group compared to the 270 COV+ non-severe group ( Figure 3J , Table S3 ). Diabetes, cancer, and chronic kidney diseases To give further confidence that our predictor metabolites are associated with COVID-19 324 pathogenesis, we aimed to determine how the levels of these metabolites changed over the 325 course of disease progression. First, we considered the portion of the COV+ severe cohort that 326 survived SARS-CoV-2 infection. We sought to determine the temporal behavior of their 327 metabolic profiles as patients reach peak disease severity and after recovery. Accordingly, we 328 compared the longitudinal metabolite abundances from individuals who had severe disease but 329 survived and were discharged from the hospital. We compared their initial d0 plasma sample to 330 the sample taken closest to the day of ICU admission, when the disease had progressed to peak 331 severity. We also compared their initial d0 plasma sample to the last sample provided by the 332 patient at or after hospital discharge. For several LPCs and one PC, a V-shaped trend was 333 observed ( Figure 4A ). After the initial sample (d0), the level of these metabolites dropped further 334 as the disease worsened but then began to return to d0 levels during recovery. The reverse trend 335 was observed for Cer-NS d18:1_16:0. Its levels significantly increased until the patients were 336 admitted to the ICU, and the levels dropped to below the initial d0 levels in the final sample 337 obtained. 338 These pronounced longitudinal trends in surviving COV+ severe patients raised the question 339 of how the trajectory of disease progression (as marked by our predictor metabolites) differed 340 among COV+ non-severe patients, surviving COV+ severe patients, and deceased COV+ severe 341 patients. We also wished to compare the end points in these groups to the COV-d0 patients. We Notably, the analysis revealed three distinct trajectories with starting points that trended with 349 disease severity. The groups then followed a common trajectory towards the COV-d0 sample. 350 However, the COV+ non-severe patients recovered much quicker than the COV+ severe were significantly different between d4 and d14 in these animals ( Figure S11B ). In SARS-CoV-2 416 infected hamsters, plasma LPC levels decreased on d4 compared to d2 and slowly recovered 417 towards the control levels on d14. By comparison, for the influenza-infected animals, metabolite 418 levels approached those of the control group more rapidly ( Figure 6D and Figure S11C ) forest), we found worse cross-validated performance than ElasticNet (see Figure S4C ). 447 Interpretation of our model led us to identify metabolites that predicted disease severity. 448 Using a reduced predictor set, we were able to retrain our model and found similarly strong 449 predictive ability. Of the predictor metabolites, the majority were LPCs and PCs that decreased sample in each batch. The QC sample was injected after every 12 th sample. After peak area 773 extraction, batch effects were observed in the research samples (see Figure S2A ). The research 774 samples and QC data were used to test typical batch normalization methods (see Figure S2B ), 775 including constant sum, unit length, scale, percentile shift, minimum-maximum, PQN, quantile 776 and ComBat correction used in metabolomics 29,72-75 . In Figure S2B , the variance remaining in 777 the research samples normalized to the variance in the QC samples is shown for each method. 778 The higher this ratio, the more variance that remains in the research samples and the more batch-779 J o u r n a l P r e -p r o o f derived variance in the QC samples is reduced. ComBat correction outperformed the other batch-780 correction approaches tested by using this metric. After correction, samples are well clustered 781 according to sample type (WU-350, QC, blank) as shown in Figure S2C . In addition, within the 782 research samples, there is no clustering by batch (see Figure S2D ). To control for disease severity in the correlation analysis of the predictor metabolites, 

applying 2x 500 µL 1:1 methyl tert-butyl ether:methanol (v/v) onto the SPE cartridge and 712 centrifuging for 2 min at 1000 g and 4 C. The combined eluates were dried under a stream of 713 nitrogen (Biotage SPE Dry Evaporation System) at room temperature and reconstituted

Hamster plasma samples were diluted 1:4 with methanol (v/v), vortexed for 30 seconds, and 716 incubated at -20C for 2 hours. Samples were centrifuged for 10 minutes at 13,500 x g at 4°C 717 and supernatant was transferred to a new centrifuge tube, concentrated, and stored at -80C until 718 reconstitution as described above

LC/MS analysis of polar metabolites

An aliquot of 2 µL of polar metabolite extract was subjected to LC/MS analysis by using an 721 Agilent 1290 Infinity II liquid-chromatography (LC) system coupled to an Agilent 6540

Polar metabolites were separated on a SeQuant® ZIC®-pHILIC 724 column (100 x 2.1 mm, 5 µm, polymer, Merck-Millipore) including a ZIC®-pHILIC guard 725 column (2.1 mm x 20 mm, 5 µm)

20 mM ammonium bicarbonate, 0.1% ammonium hydroxide solution (25% 728 ammonia in water), 2.5 µM medronic acid, and B: 95% acetonitrile, 5% water, 2.5 µM medronic 729 acid. The following linear gradient was applied: 0 to 1 min

90% B followed by a re-equilibration phase of 4 min at 400 µLmin -1 and 2 731 min at 250 µLmin -1 . Metabolites were detected in positive and negative ion mode with the 732 following source parameters: gas temperature 200 C, drying gas flow 10 Lmin -1 , nebulizer , sheath gas temperature 300C, sheath gas flow 12 Lmin -1 , VCap 3000 V, nozzle 734 voltage 2000 V, Fragmentor 100 V, Skimmer 65 V, Oct 1 RF Vpp 750 V, and m/z range 50-735 1700. Data were acquired under continuous reference mass correction at m/z 121.0509 and 736 922.0890 for positive ion mode and m/z 119.0363 and 966.0007 for negative ion mode. Samples 737 were randomized prior to analysis

LC/MS analysis of lipid metabolites

An aliquot of 2 µL of lipid extract was subjected to LC/MS analysis by using an Agilent 741

Infinity II LC-system coupled to an Agilent 6545 Q-TOF mass spectrometer with a dual 742

Agilent Jet Stream electrospray ionization source

HSS T3 column (2.1 x 150 mm, 1.8 µm) including an Acquity UPLC® HSS T3 VanGuard Pre

at a temperature of 60 C and a flow rate of 250 µLmin -1 . The 745 mobile phases consisted of A: 60% acetonitrile, 40% water, 0.1% formic acid, 10 mM 746 ammonium formate, 2.5 µM medronic acid, and B: 90% 2-propanol, 10% acetonitrile, 0.1% 747 formic acid, 10 mM ammonium formate (dissolved in 1 mL water). The following linear gradient 748 was used: 0-2 min, 30% B; 17 min

Lipids were detected in positive and negative ion mode with the following source parameters: 751 gas temperature 250 C, drying gas flow 11 Lmin -1 , nebulizer pressure 35 psi, sheath gas 752 temperature 300 C, sheath gas flow 12 Lmin -1

Skimmer 65 V, Oct 1 RF Vpp 750 V, and m/z range 50-1700. Data were 754 acquired under continuous reference mass correction at m/z 121.0509 and 922.0890 in positive 755 ion mode and m/z 119.0363 and 966.0007 in negative ion mode

A new coronavirus associated with human respiratory disease in China

Wider Image in COVID-hit India, a 26-year-old-doctor decides who lives and who dies

Therapy for Early COVID-19: A Critical Need

Potency and timing of antiviral therapy as 901 determinants of duration of SARS-CoV-2 shedding and intensity of inflammatory response

An Algorithm for Classifying Patients Most Likely to Develop Severe 905 Coronavirus Disease

Angiotensin-converting enzyme 2 907 (ACE2) as a SARS-CoV-2 receptor: molecular mechanisms and potential therapeutic target

SARS-CoV-2 Reverse Genetics Reveals a Variable Infection 911 Gradient in the Respiratory Tract

Patients: Identification of Diagnostic and Prognostic Biomarkers

COVID-19 infection alters kynurenine and fatty acid 918 metabolism, correlating with IL-6 levels and renal status

Proteomic and Metabolomic Characterization of COVID-19 Patient Sera

Plasma metabolomic and lipidomic alterations associated with COVID-19

Integrative Modeling of Quantitative Plasma Lipoprotein

Acid Data Reveals a Multiorgan Pathological Signature of SARS-CoV-2 Infection

The 933 serum metabolome of COVID-19 patients is distinctive and predictive

Understanding the pathophysiological changes via untargeted metabolomics 937 in COVID-19 patients

Metabolomic/lipidomic profiling of COVID-19 and individual response to 940 tocilizumab

Covid-19 Automated Diagnosis and Risk 943 Assessment through Metabolomics and Machine Learning

High-throughput approaches of diagnosis and therapies for COVID-19: antibody panels, 946 proteomics and metabolomics

Innovation: Metabolomics: the apogee of the omics trilogy

Metabolomics enables precision medicine

Animal models for COVID-19. 955

Simulation of the Clinical and Pathological Manifestations of Coronavirus Disease 958 2019 (COVID-19) in a Golden Syrian Hamster Model: Implications for Disease Pathogenesis and 959 Transmissibility

Syrian hamsters as a small animal model for SARS-CoV-2 962 infection and countermeasure development

Clinical Characteristics of Covid-19 in New York City

Coronavirus Disease 2019 (COVID-19) -SymptomsCenters for Disease Control and 968 Prevention

Comorbidity Measures for Use with 970 Administrative Data

Intensity drift removal in LC/MS metabolomics by common variance compensation

Systems biological assessment of immunity to mild versus 976 severe COVID-19 infection in humans

Factors associated with hospital admission and critical illness 979 among 5279 people with coronavirus disease 2019 in New York City: prospective cohort study

Omics-Driven Systems Interrogation of Metabolic Dysregulation in COVID-19

Metabolomics to Predict Antiviral Drug Efficacy in COVID-19

Regularization and Variable Selection via the Elastic Net

Migrating from partial least squares discriminant 990 analysis to artificial neural networks: a comparison of functionally equivalent visualisation and feature 991 contribution tools using jupyter notebooks

Integrative 993 clinical, genomics and metabolomics data analysis for mainstream precision medicine to investigate 994 COVID-19

Prognostic Value of C-Reactive Protein in Patients With Coronavirus

A pneumonia outbreak associated with a new coronavirus of probable bat origin

COVID-19: 1002 immunopathogenesis and Immunotherapeutics

Re-analysis of Single Cell Transcriptome Reveals That the NR3C1-1005 CXCL8-Neutrophil Axis Determines the Severity of COVID-19

Single-cell landscape of bronchoalveolar immune cells in patients with COVID-19

Hematological findings and 1012 complications of COVID-19

Lymphopenia predicts disease severity of COVID-19: a descriptive and predictive study

Lymphopenia predicted illness severity and recovery in patients with COVID-19: A single-center, 1018 retrospective study

Lymphopenia an important 1020 immunological abnormality in patients with COVID-19: Possible mechanisms

Lymphopenia in COVID-19: Therapeutic opportunities

Impaired glucose metabolism in patients with diabetes, prediabetes, and 1026 obesity is associated with severe COVID-19

Predictive symptoms and comorbidities for severe COVID-19 and 1028 intensive care unit admission: a systematic review and meta-analysis

Noise Reduction in Speech Processing

Correlation Between a Discrete and a Continuous Variable. Point-Biserial Correlation

Exploring blood 1035 alterations in chronic kidney disease and haemodialysis using metabolomics

Distinct inflammatory profiles distinguish COVID-19 from 1039 influenza with limited contributions from cytokine storm

Interleukin-6 Is a Biomarker for the Development of Fatal Severe 1042

Acute Respiratory Syndrome Coronavirus 2 Pneumonia

An inflammatory cytokine signature predicts COVID-1046 19 severity and survival

Interleukine-6 in critically ill COVID-19 patients: A 1049 retrospective analysis

Syrian Hamster as an Animal Model for the Study of Human Influenza Virus 1052 Infection

TOP1 inhibition therapy protects against SARS-CoV-2-1055 induced lethal inflammation

Fact sheet for health care providers emergency use authorization (EUA) of casirivimab and 1057 imdevimab

A comparative evaluation of the generalised 1059 predictive ability of eight machine learning algorithms across ten clinical metabolomics data sets for binary 1060 classification

Large-Scale Plasma Analysis Revealed New Mechanisms and 1063 Molecules Associated with the Host Response to SARS-CoV-2

The COVID-19 Cytokine 1065 Storm; What We Know So Far

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Decreased plasma phospholipid concentrations and increased 1074 acid sphingomyelinase activity are accurate biomarkers for community-acquired pneumonia

Plasma ceramide and lysophosphatidylcholine inversely correlate with mortality in sepsis patients

Coronavirus biology and 1080 replication: implications for SARS-CoV-2

SARS-CoV-2 infection rewires host cell 1083 metabolism and is potentially susceptible to mTORC1 inhibition

Favor SARS-CoV-2 Infection and Monocyte Response through a HIF-1alpha/Glycolysis-Dependent Axis

Genetic Screens Identify Host 1091 Factors for SARS-CoV-2 and Common Cold Coronaviruses

HMDB 4.0: the human metabolome database for

MassBank: a public repository for sharing mass spectral data for life sciences

Statistical Analysis of Proteomics, Metabolomics, and Lipidomics Data Using Mass 1100

Mertens Springer Frontiers in Probability and the Statistical 1101

hardcover ISBN

Adjusting batch effects in microarray expression data 1104 using empirical Bayes methods

Non-targeted UHPLC-MS metabolomic data processing methods: a comparative 1107 investigation of normalisation, missing value imputation, transformation and scaling

0: towards more transparent and integrative metabolomics analysis

Applications of machine learning and 1113 artificial intelligence for Covid-19 (SARS-CoV-2) pandemic: A review

Author Correction: SciPy 1.0: fundamental algorithms 1117 for scientific computing in Python

Scikit-learn: Machine Learning in Python

Structural Determination of Lysophospholipid Regioisomers by 1122

Electrospray Ionization Tandem Mass Spectrometry †

Quantitative lysophospholipidomics in 1125 human plasma and skin by LC-MS/MS

Separation and quantification of 2-acyl-1-lysophospholipids and 1-acyl-2-1128 lysophospholipids in biological samples by LC-MS/MS

SciPy 1.0: fundamental algorithms for scientific 1131 computing in Python

Controlling the False Discovery Rate: A Practical and Powerful 1134 Approach to Multiple Testing