Antibodybased measurements of the HUPO reference specimens

The PPP reference specimens were distributed to four different laboratories for immunoassay or antibody microarray analysis. Each of the four sites used a distinct technology for analyzing the specimens. The 39 immunoassays performed on DB clinical analyzers were based on immunonephelometric methods (33 tests) and sandwich-like enzyme immunoassays (6 tests) that use antibody-coated magnetic chromium dioxide particles. The GNF measured 88 different serum proteins using microarray-based sandwich assays detected by RLS. MSI used antibody micro-arrays to target 168 different proteins, mostly cytokines, using sandwich assays and detection by RCA [4, 21]. VARI measured 28 different serum proteins using TC-RCA detection on antibody microarrays [19]. The antibodies used by each site are listed in Supplemental Tab. 1. Each site independently designed their own experiments based on individual resources and experience, and the targeted proteins varied significantly between sites. The complete data sets are available at http:// www.vai.org/vari/labs/haab.asp.

Two of the sites (MSI and GNF) ran the samples in triplicate, one in duplicate (VARI), and one had duplicate measurements for four of the samples (DB). The reproducibility of the replicate data is a good indicator of data quality. Replicate measurements showed good reproducibility for each data set, as depicted by the correlations of the different antibody measurements for the same sample in two separate experiments (Fig. 1). The average correlation coefficients between the different antibody measurements from replicate experiments were 0.99 for the DB set, 0.95 for the GNF set, 0.94 for the MSI set, and 0.96 for the VARI set. These high average correlations indicate that each data set is highly internally consistent.

Fig. 1 Correlations between replicate measurements of one sample. Duplicate antibody measurements from a plasma sample were plotted against each other. Scatter plots from each of the four data sets are shown: (A) DB; (B) GNF; (C) MSI; (D) VARI. Correlations for each of the plots were 0.99, 0.95, 0.94, and 0.96, respectively. Plots are shown for the samples: (A) CAMS, citrate-plasma; (B) BDAA, citrate-plasma; (C) BDAF, citrate-plasma; and (D) BDCA, heparin-plasma.

Fig. 1 Correlations between replicate measurements of one sample. Duplicate antibody measurements from a plasma sample were plotted against each other. Scatter plots from each of the four data sets are shown: (A) DB; (B) GNF; (C) MSI; (D) VARI. Correlations for each of the plots were 0.99, 0.95, 0.94, and 0.96, respectively. Plots are shown for the samples: (A) CAMS, citrate-plasma; (B) BDAA, citrate-plasma; (C) BDAF, citrate-plasma; and (D) BDCA, heparin-plasma.

Two of the data sets (GNF, MSI) used standard curves of purified antigens to calibrate the data and to calculate the concentrations of each of the measured proteins. DB analyzers used reference materials (standards, controls, and calibrators) that are based on IRMs and purified antigens for calibration and the determination of the concentrations of the analytes. The measured concentrations cover a broad range, from several mg/mLto below 1 pg/mL (Fig. 2). The GNF and MSI data sets, focusing on cytokine detection, account for most of the low-abundance measurements, while the DB and VARI sets focused on common mid-to-high-abundance serum proteins. Some overlap existed between the sets: six analytes were common between DB and GNF, three were common between DB and MSI, 11 were common between DB and VARI, 57 were common between GNF and MSI, 10 were common between GNF and VARI, and nine were common between MSI and VARI.

While the precision between replicates within each data set is good (Fig. 1), occasionally large differences were observed between platforms in the measured concentrations of common analytes. Of the 57 common analytes between GNF and

Fig. 2 Concentration range of the proteins measured in these studies. Geometric mean concentration over all the samples is plotted for each of 295 quantitative assays for 231 unique proteins. Set consists of quantitative measurements from the DB BN system (33 analytes), the DB Dimension system (6 analytes), MSI antibody microarrays (168 analytes), and GNF antibody microarrays (88 analytes). For analytes measured by more than one laboratory, the geometric mean concentration derived by each laboratory is displayed.

Fig. 2 Concentration range of the proteins measured in these studies. Geometric mean concentration over all the samples is plotted for each of 295 quantitative assays for 231 unique proteins. Set consists of quantitative measurements from the DB BN system (33 analytes), the DB Dimension system (6 analytes), MSI antibody microarrays (168 analytes), and GNF antibody microarrays (88 analytes). For analytes measured by more than one laboratory, the geometric mean concentration derived by each laboratory is displayed.

MSI, seven were measured more than ten-fold higher at GNF and eight were measured more than ten-fold higher at MSI. These deviations between assays in the measurement of common analytes can be seen in Fig. 2. Supplemental Tab. 2 provides the average measured concentrations of the analytes that were measured at more than one site. Interlaboratory variation is not uncommon and may be due to differences in the specificities of the antibodies used, the sample storage and treatment methods, and the calibration methods. The full exploration of the sources of variation between the laboratories was beyond the scope of this study, yet the existence of the occasional variation highlights the need for methods for calibration and validation across laboratories and platforms.

Systematic variation between the preparation methods of the PPP reference specimens

We investigated whether the blood preparation methods (serum, citrate-plasma, EDTA-plasma, heparin-plasma) introduced systematic bias into the abundances of all the proteins in general. A systematic bias in concentration would be evidenced by a consistent shift in the concentrations of analytes in one preparation method relative to the other methods. The protein abundances were compared between the samples that were prepared from the same starting material, i.e., we compared the four preparations within the BDAA specimen set, the four preparations within the

Tab. 2 Concentrations and associated MS summary information. Information relating to 70 I PI numbers (66 unique analytes) that had a match between the analyte-derived lists and the MS-derived lists is presented. "Antibody name" = the name that was used in the searches for analyte-associated I PI numbers. "Name from analyte search" = the name in the I PI database that matched the antibody/ analyte name. "Concentration" = the geometric mean concentration over all specimens as found by immunoassay or antibody microarray. "# Labs" = the number of laboratories (out of 18) that found a particular I PI number. "# Peptides" = the average number of different peptides found for that I PI number. "I PI set" = the analyte-associated list from which a match was found (see Section 2), either list 1, list 2, or both (1, 2, or B). In four instances, two different IPI numbers were associated with one analyte

Antibody name

Name from analyte search

Concentration,

#

#

Laboratory

IPI

pg/mL

Labs

Peptides

SET

Albumin

Albumin

4.0E + 10

17

201

DB

2

Transferrin

Transferrin

2.3E + 09

16

249

DB

2

Apolipoprotein A I

Apolipoprotein A-I

1.4E + 09

17

82

DB

2

a2-macroglobulin

Alpha-2-macroglobulin

1.4E + 09

17

211

DB

2

al-antitrypsin

Serine (or cysteine) proteinase inhibitor, clade A (alpha-1 antiproteinase, antitiypsin), member 1

1.1E + 09

15

183

DB

2

C3c

Complement component 3

9.5E + 08

5

98

DB

2

Haptoglobin

Haptoglobin

8.8E + 08

18

113

DB

2

Hemopexin

Hemopexin

7.5E + 08

16

86

DB

2

Apolipoprotein B

Apolipoprotein B (including Ag(x) antigen)

7.2E + 08

13

328

DB

2

Fibrinogen

Fibrinogen, gamma polypeptide

6.7E + 08

16

66

DB

2

Fibrinogen

Fibrinogen, gamma polypeptide

6.7E + 08

12

51

DB

2

al-acid-glycoprotein

Alpha-l-acid glycoprotein 2 precursor

6.IE + 08

14

24

DB

1

al-acid-glycoprotein

Orosomucoid 1

6.IE + 08

16

45

DB

2

Antithrombin III

Serine (or cysteine) proteinase inhibitor, clade C (antithrombin), member 1

3.2E + 08

17

70

DB

2

Apolipoprotein A-II

Apolipoprotein A-II

3.0E + 08

15

18

DB

2

Prealbumin

Transthyretin (prealbumin, amyloidosis type I)

2.6E + 08

17

27

DB

2

Cemloplasmin

Cemloplasmin (ferroxidase)

2. IE + 08

15

134

DB

2

Antibody name

Name from analyte search

Concentration,

#

#

Laboratory

IPI

pg/mL

Labs

Peptides

SET

C4

Complement C4 precursor [Contains: C4A anaphylatoxin]

1.7E + 08

17

157

DB

1

Plasminogen

Plasminogen

1.4E + 08

12

72

DB

2

Fibronectin

Fibronectin 1

1.1E + 08

1

86

DB

2

Apolipoprotein E

Apolipoprotein E

3.4E + 07

8

30

DB

2

vWF

Von Willebrand factor

1.3E + 06

2

46

GNF

2

P2Microglobulin

Beta 2-microglobulin protein

1.1E + 06

1

1

DB

1

P2Microglobulin

Beta-2-microglobulin

1.1E + 06

3

1

DB

2

sTfR

Transfenin receptor (p90, CD71)

5.8E + 05

1

2

DB

2

VAP-1

Amine oxidase, copper containing 3 (vascular adhesion protein 1)

1.2E + 05

2

6

MSI

2

Protein C

Mannose-binding lectin (protein C) 2, soluble (opsonic defect)

9.7E + 04

2

7

MSI

2

VCAM-I

Vascular cell adhesion molecule 1

9.4E + 04

3

9

MSI/GNF

2

TGFpi

Transforming growth factor, beta 1 (Camurati-Engelmann disease)

7.5E + 04

2

2

GNF

2

IGF-BP3

Insulin-like growth factor binding protein 3

5.9E + 04

6

17

MSI/GNF

2

ICAM-1

Intercellular adhesion molecule 1 (CD54), human rhinovims receptor

4.3E + 04

2

4

MSI/GNF

2

MMPg

Matrix metalloproteinase 9 (gelatinase B, 92 kDa gelatinase, 92 kDa type IV collagenase)

4.1E + 04

2

5

MSI/GNF

2

VE-cadherin

Cadherin 5, type 2, VE-cadherin (vascular epithelium)

3.0E + 04

3

11

MSI

2

M-CSF R

Colony stimulating factor 1 receptor, formerly McDonough feline sarcoma viral (v-fms) oncogene homolog

2.6E + 04

3

11

MSI

2

L-Selectin

Selectin L (lymphocyte adhesion molecule 1)

1.7E + 04

5

10

MSI

2

ALCAM

Activated leukocyte cell adhesion molecule

1.6E + 04

2

5

MSI

2

IGFBP2

Insulin-like growth factor binding protein 2, 36 kDa

1.5E + 04

1

3

MSI

2

TIMP1

Tissue inhibitor of metalloproteinase 1

(erythroid potentiating activity, collagenase inhibitor)

1.4E + 04

1

3

MSI/GNF

2

Tab. 2 Continued

Antibody name

Name from analyte search

Concentration, # pg/mL Labs

Peptides

Laboratory

IPI SET

EGF R1

Epidermal growth factor receptor (erythroblastic leukemia viral (v-erb-b) oncogene homolog, avian)

1.1E + 04 3

3

GNF

2

MMP2

Matrix metalloproteinase 2 (gelatinase A, 72 kDa gelatinase, 72 kDa type IV collagenase)

8.8E + 03 1

7

MSI/GNF

2

NAP-2

Nucleosome assembly protein 1-like 4

7.5E + 03 1

1

MSI

2

LIF Ra

Leukemia inhibitoiy factor receptor

5.0E + 03 2

4

MSI

2

PDGF-Ra

Platelet-derived growth factor receptor, alpha polypeptide

4.6E + 03 1

2

MSI

2

MMP1

Matrix metalloproteinase 1 (interstitial collagenase)

2.6E + 03 1

1

MSI/GNF

2

FasL

Tumor necrosis factor (ligand) superfamily, member 6

1.5E+03 1

2

MSI/GNF

2

NSE

Enolase 2 (gamma, neuronal)

1.4E + 03 1

1

GNF

2

MMP8

Matrix metalloproteinase 8 (neutrophil collagenase)

9.0E + 02 1

1

MSI/GNF

2

VEGF-D

C-fos induced growth factor (vascular endothelial growth factor D)

5.0E + 02 1

1

MSI/GNF

2

ENA-78

Chemokine (C-X-C motif) ligand 5

3.4E + 02 1

1

MSI

2

CD30

Tumor necrosis factor receptor superfamily, member 8

3.3E + 02 1

2

MSI/GNF

2

MPIF-1

Chemokine (C-C motif) ligand 23

3.2E + 02 1

1

MSI

2

GROb

Chemokine (C-X-C motif) ligand 2

3.0E + 02 1

1

MSI

2

BDNF

Brain-derived neurotrophic factor

3.0E + 02 1

1

MSI

2

AFP

Alpha-fetoprotein

2.9E + 02 2

2

MSI/GNF

2

IGF-IR

Insulin-like growth factor 1 receptor

2.4E + 02 1

1

MSI/GNF

2

Calcitonin

Calcitonin/calcitonin-related polypeptide, alpha

1.9E + 02 1

1

GNF

2

Calcitonin

Calcitonin gene-related peptide type 1 receptor precursor

1.9E + 02 1

1

GNF

1

FGFß

Fibroblast growth factor-20

1.6E + 02 1

1

GNF

1

IL-10Rß

Interleukin 10 receptor, beta

1.5E + 02 1

1

MSI

2

Angp2

Angiopoietin 2

9.7E + 01 1

1

GNF

2

MCP-1

Splice isoform A of P15529

7.6E + 01 1

1

MSI/GNF

1

Antibody name

Name from analyte search

Concentration,

#

# Laboratory

IPI

pg/mL

Labs

Peptides

SET

SCF

KIT ligand

5.9E + 01

1

1 MSI/GNF

2

IFNy

Interferon, gamma

5.4E + 01

1

1 MSI/GNF

2

OSM

Oncostatin M

4.8E + 01

1

1 MSI

2

ILla

Interleukin 1, alpha

4.5E + 01

1

1 MSI/GNF

2

TNFa

Tumor necrosis factor (TNF superfamily, member 2)

3.7E + 01

2

1 MSI/GNF

2

AR

Androgen receptor (dihydrotestosterone receptor; testicular feminization; spinal and bulbar muscular atrophy; Kennedy disease)

2.6E + 01

1

1 MSI/GNF

2

I-TAC

Chemokine (C-X-C motif) ligand 11

2.3E + 01

1

1 MSI

2

CGB

Chorionic gonadotropin, beta polypeptide

1.9E + 01

1

1 MSI/GNF

2

IL7

Interleukin 7

7.0E + 00

1

1 MSI/GNF

2

BDAF specimen set, etc. For each preparation type (citrate-plasma, EDTA-plasma, etc.), the number of proteins that had a maximum concentration in that preparation was totaled. The number of proteins with minimum concentrations also was totaled for each preparation method. Those numbers were compared to the numbers of maxima or minima that would be expected by chance. Frequencies of maxima or minima much greater or lower than would be expected by chance could indicate systematic bias in the concentrations in a particular preparation method.

The results of that analysis are shown in Fig. 3. The proportion of proteins that had maxima (Fig. 3A) or minima (Fig. 3B) in each preparation type is indicated by the position on the x-axis of a different vertical line for each of the four data sets. The distribution of maxima and minima in each preparation method that would be expected by chance was calculated by permutation and is indicated by the histograms in each plot. As expected, the average frequency in the randomly permuted data is 0.25, since the maxima and minima are evenly distributed among the four preparation methods. All four data sets had a significantly lower frequency of maxima in citrate-plasma (Fig. 3A, top left), well below what is expected by chance. The GNF and MSI sets showed a high frequency of maxima in the EDTA-plasma samples (Fig. 3A, top right); the VARI measurements were often highest in heparin-plasma (Fig. 3A, lower left), and the DB measurements were frequently highest in the serum samples (Fig. 3A, lower right). For the minimum values, all methods showed a significantly frequent occurrence of minima for the citrate samples (Fig. 3B, top left), and the DB data were very seldom lowest using heparin-plasma or serum samples (Fig. 3B, lower left and lower right). The other frequencies are close to what might be expected by chance. These analyses show evidence for general biases in protein concentrations as a result of blood preparation method.

We examined the magnitudes of concentration differences between the sample types. For each protein, the concentration in each preparation method was divided by the maximum concentration found in that specimen set. For example, if a protein had a concentration of 100 pg/mL in citrate-plasma and 200 pg/mL in serum, citrate-plasma was given a 0.5 and serum was given a 1.0. The median concentration ratios for each preparation method are shown in Fig. 4 for each of the four data sets. Each data set shows the citrate-plasma preparation with the lowest average abundances, from about 85% of the maximum values (DB) to about 40% of the values (GNF and MSI). Consistent with the results from Fig. 3, serum had the highest concentrations in the DB set, EDTA-plasma in the MSI and GNF sets, and heparin-plasma in the VARI set. The variation between preparation methods is similar between the DB and VARI sets and between the GNF and MSI sets, and the GNF and MSI sets had broader variation in the relative abundances (larger error bars) than the other two sets. These relationships could be related to the similarity between the groups in the proteins measured; GNF and MSI measured mostly cytokines, while VARI and DB measured higher-abundance serum proteins.

Fig. 3 Frequency of maxima (A) and mini- preparation method. Each line represents one ma (B) in each ofthe four preparation methods: of the four data sets, with the identities given in citrate-plasma (upper left), EDTA-plasma the legend: DB = dashed line, GNF = dotted/

(upper right), heparin-plasma (lower left), and dashed line, MSI = solid line, VARI = dotted serum (lower right). Position on the x-axis for line. Distribution of randomly-occurring fre-

the vertical lines in each plot indicate the fre- quencies also is included in each graph. quency of maxima (A) or minima (B) in a given

Fig. 3 Frequency of maxima (A) and mini- preparation method. Each line represents one ma (B) in each ofthe four preparation methods: of the four data sets, with the identities given in citrate-plasma (upper left), EDTA-plasma the legend: DB = dashed line, GNF = dotted/

(upper right), heparin-plasma (lower left), and dashed line, MSI = solid line, VARI = dotted serum (lower right). Position on the x-axis for line. Distribution of randomly-occurring fre-

the vertical lines in each plot indicate the fre- quencies also is included in each graph. quency of maxima (A) or minima (B) in a given

Fig. 4 Median relative concentration ratios in each sample type. Concentration of each protein in each preparation type was divided by the concentration of the preparation type that was highest for a given sample. Median relative change in concentration is depicted for each preparation type from the (A) DB, (B) GNF, (C) MSI, and (D) VARI data sets. Error bars represent the SD in relative concentration change over all the proteins.

Fig. 4 Median relative concentration ratios in each sample type. Concentration of each protein in each preparation type was divided by the concentration of the preparation type that was highest for a given sample. Median relative change in concentration is depicted for each preparation type from the (A) DB, (B) GNF, (C) MSI, and (D) VARI data sets. Error bars represent the SD in relative concentration change over all the proteins.

Diabetes 2

Diabetes 2

Diabetes is a disease that affects the way your body uses food. Normally, your body converts sugars, starches and other foods into a form of sugar called glucose. Your body uses glucose for fuel. The cells receive the glucose through the bloodstream. They then use insulin a hormone made by the pancreas to absorb the glucose, convert it into energy, and either use it or store it for later use. Learn more...

Get My Free Ebook


Post a comment