WO2009156747A2 - Détermination - Google Patents

Détermination Download PDF

Info

Publication number
WO2009156747A2
WO2009156747A2 PCT/GB2009/001624 GB2009001624W WO2009156747A2 WO 2009156747 A2 WO2009156747 A2 WO 2009156747A2 GB 2009001624 W GB2009001624 W GB 2009001624W WO 2009156747 A2 WO2009156747 A2 WO 2009156747A2
Authority
WO
WIPO (PCT)
Prior art keywords
fingerprint
precursor
sample
subject
ions
Prior art date
Application number
PCT/GB2009/001624
Other languages
English (en)
Other versions
WO2009156747A3 (fr
Inventor
Robert Charles Rees
Graham Roy Ball
Balwir Matharoo-Ball
Kevin Morgan
Noor Ahmed Kalshekar
Original Assignee
The Nottingham Trent University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by The Nottingham Trent University filed Critical The Nottingham Trent University
Publication of WO2009156747A2 publication Critical patent/WO2009156747A2/fr
Publication of WO2009156747A3 publication Critical patent/WO2009156747A3/fr

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/48—Biological material, e.g. blood, urine; Haemocytometers
    • G01N33/50—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing
    • G01N33/68—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing involving proteins, peptides or amino acids
    • G01N33/6803—General methods of protein analysis not limited to specific proteins or families of proteins
    • G01N33/6848—Methods of protein analysis involving mass spectrometry
    • G01N33/6851—Methods of protein analysis involving laser desorption ionisation mass spectrometry
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/48—Biological material, e.g. blood, urine; Haemocytometers
    • G01N33/50—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing
    • G01N33/68—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing involving proteins, peptides or amino acids
    • G01N33/6893—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing involving proteins, peptides or amino acids related to diseases not provided for elsewhere
    • G01N33/6896—Neurological disorders, e.g. Alzheimer's disease
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2333/00—Assays involving biological materials from specific organisms or of a specific nature
    • G01N2333/435—Assays involving biological materials from specific organisms or of a specific nature from animals; from humans
    • G01N2333/46—Assays involving biological materials from specific organisms or of a specific nature from animals; from humans from vertebrates
    • G01N2333/47—Assays involving proteins of known structure or function as defined in the subgroups
    • G01N2333/4701—Details
    • G01N2333/4709—Amyloid plaque core protein
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2800/00—Detection or diagnosis of diseases
    • G01N2800/28—Neurological disorders
    • G01N2800/2814—Dementia; Cognitive disorders
    • G01N2800/2821—Alzheimer
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N30/00—Investigating or analysing materials by separation into components using adsorption, absorption or similar phenomena or using ion-exchange, e.g. chromatography or field flow fractionation
    • G01N30/02—Column chromatography
    • G01N30/62—Detectors specially adapted therefor
    • G01N30/72—Mass spectrometers
    • G01N30/7233—Mass spectrometers interfaced to liquid or supercritical fluid chromatograph

Definitions

  • the invention relates to a mass spectrometry fingerprint for identifying a dementia brain disorders such as Alzheimer's disease (AD) or mild cognitive impairment (MCI), to the use of that fingerprint to identify such a brain disease, or a stage of that disease, and to peptide markers and their diagnostic uses, resulting from that fingerprint are also the subject of this application.
  • AD Alzheimer's disease
  • MCI mild cognitive impairment
  • AD Alzheimer's Disease
  • pathological hallmarks including senile plaques mainly composed of aggregated amyloid-beta (A ⁇ ) and neurofibrillary tangles (NFT) of hyperphosphorylated tau protein. It can be diagnosed in most clinics with over 90% accuracy using tests of cognitive function. However, this is only at a late stage of the disease when there is extensive irreversible neuronal damage already present.
  • AD pathology often accumulates over a 10 to 20 year period before clinical symptoms present and even patients with mild cognitive impairment (MCI) have been found to display marked AD pathology at PM.
  • MCI mild cognitive impairment
  • a reliable, accurate biomarker or set of biomarkers would be an invaluable aid to clinical diagnosis and potentially identify affected individuals much earlier when drug intervention is likely to be most effective.
  • biomarkers to aid clinical diagnosis and progression of Alzheimer's disease (AD).
  • Recent advancements in proteomic technologies are facilitating high throughput analysis of protein expression patterns 1 .
  • This in conjunction with specific statistical algorithms can be used to discriminate samples according to their disease status 3"5 .
  • Studies have shown that the accuracy of the clinical diagnosis is between 65 and 90% using cognitive techniques. Higher accuracy is achieved at academic centres with special interest in AD. However, this is only at a late stage of the disease when there is extensive irreversible neuronal damage already present.
  • AD pathology often accumulates over a 10 to 20 year period before clinical symptoms present and even patients with mild cognitive impairment (MCI) have been found to display marked AD pathology.
  • MCI mild cognitive impairment
  • the accuracy of the clinical diagnoses at the primary care level, and in general hospitals are probably even lower, especially in the early stages of the disease when the symptoms are indistinct.
  • a reliable, accurate biomarker or set of biomarkers would be an invaluable aid to clinical diagnosis and potentially identify affected individuals much earlier when drug intervention is likely to be most effective.
  • CSF cerebrospinal fluid
  • the protein is normal for 1-2 days, but thereafter, there is a marked increase after 2-3 weeks (Hesse et al., 2001), which reflects the leakage of tau from damaged neurons thus lending support to the idea that biochemical analyses does indeed reflect pathogenic processes of the brain in the CSF.
  • biochemical analyses does indeed reflect pathogenic processes of the brain in the CSF.
  • the inventors have identified some markers for AD in CSF.
  • MALDI/TOF-MS time-of-flight mass spectrometry
  • MALDI mass spectrometry has become established as the preferred choice for the rapid analyses biological samples. This is due in part to the fact that MALDI ionisation is less susceptible to ion suppression effects that are often seen with other sources (e.g. electrospray). The limits of detection for MALDI (sub-femtomole) allow small quantities of samples to be analysed and this is of crucial importance especially where clinical material is concerned.
  • ANNs Artificial Neural Networks
  • ANNs are a form of machine learning algorithm capable of accurately modeling biological systems and identifying biomarkers. They are capable of producing models that discriminate between multiple classes for blind data 3 , ' while allowing parameterization to determine the influence of causal agents within the system.
  • ANNs have been applied to a number of diverse areas for the identification of "biologically relevant" molecules, including pyrolysis mass spectrometry 14 and genomic micro-arraying of tumour tissue 15 .
  • MLP multi-layer perceptron
  • the inventors have developed a MALDI-MS proteomic methodology utilizing ANNs for the identification of novel biomarkers in CSF from AD patients that will aid diagnosis of the disease. It is expected that the markers are also likely to be found in plasma/serum because blood is in direct contact with the blood-brain barrier.
  • the invention provides a mass spectroscopy dementia brain condition fingerprint.
  • a mass spectroscopy dementia brain condition fingerprint for identifying the presence of a dementia brain condition and/or identifying the progression of a dementia brain condition comprising peaks derived from the following ions:
  • ions derived from a proteolytically digested peptide the ions having (m/z) values of: 2048, 1434 and 1336 (+/- 0.02%) and optionally one or more of the following additional ions, the additional ions having (m/z) values of: 2734, 1018, 3225 and/or 3121 (+/- 0.02%).
  • the term "dementia brain condition” is intended to mean any disease of the brain causing clinical symptoms of dementia. Preferred conditions include Alzheimer's Disease (AD) and Mild Cognitive Impairment (MCI). AD is especially preferred.
  • AD Alzheimer's Disease
  • MCI Mild Cognitive Impairment
  • the Fingerprint may, for example be a series of ions having the m/z values defined above or a physical representation of those ions such as a signal or a paper or electronic recording of a signal produced by the ions, such as stored on a computer memory, CD-rom or DVD-rom.
  • the fingerprint uses 4, 5, 6 or all 7 peaks.
  • the fragment comprises or has the sequence shown in the above table.
  • a further aspect of the invention provides a method of identifying a dementia brain disease comprising identifying whether a test sample obtained from a test subject suspected of having the dementia brain disease, when analysed by mass spectroscopy, shows a fingerprint according to the invention.
  • the samples used is any aspect of the invention described herein, may be obtained from any suitable tissue of fluid.
  • the fluid is cerebrospinal fluid (CSF).
  • CSF cerebrospinal fluid
  • the sample may be blood, plasma or serum.
  • the subject may be a human or a fragment or a cell obtained there from.
  • the method preferably additionally comprises obtaining a fingerprint from a control sample and comparing the fingerprint from the control sample with the fingerprint from the test sample, wherein identification of an elevation or depression in one or more ions of the fingerprint is used to determine whether the test subject has a dementia brain disease.
  • the fingerprint allows determination of whether a subject, from which a test sample has been taken, has a dementia brain disorder. For example:
  • one or more ions/peaks may be present in a control sample but absent in the test sample;
  • one or more ions/peaks may be present in a control sample but present in the test sample;
  • one or more ions/peaks may be absent in a control sample but present in the test sample
  • one or more ions/peaks may be absent in a control sample and absent in the test sample;
  • 'Absent' and 'present' may be absolute. Furthermore, these terms also refer to an increase or decrease in intensity of one or more peaks/ions necessary for the presence/absence of a dementia and/or a dementia stage to be determined.
  • proteolytic digestion is carried out using trypsin.
  • Other proteases known to the skilled person would also be suitable.
  • a chemical digestion procedure may be used.
  • the chemical digestion may result in the same fragments as a tryptic digestion.
  • a tryptic digestion is carried out under the following conditions: An aliquot (25 ⁇ l) of neat CSF may be initially fractionated using a ZipTip C 18 (Milipore, Watford, UK) with 25 cycles of binding. This may be followed by two washes in 0.1% TFA and elution in 4 ⁇ l of 80% acetonitrile/0.1 % TFA. This volume may be combined with ammonium bicarbonate (16.6 ⁇ l, 100 mM), water (7.6 ⁇ l), and trypsin Gold Mass Spectrometry Grade (1.3 ⁇ l, 0.5 ⁇ g/ ⁇ l, Promega, Victoria, UK) and incubated at 37 0 C overnight.
  • the reaction may be quenched with 1% TFA (1 ⁇ l) and the sample cleaned-up using a C 18 ZipTip following the procedure described above.
  • An aliquot of the eluate (1 ⁇ l) may be spotted onto a MALDI target using the dried droplet method with matrix, ⁇ -cyano-4- hydroxycinnamic acid (LaserBio Labs, Cedex, France), prepared as a 10 mg/ml solution in 50% acetonitrile + 0.1% TFA.
  • Duplicate samples may be applied to the target plate in a randomised order used for sample preparation and analysed by MALDI-TOF-MS. A range of appropriate blank and control samples are preferably prepared alongside the serum digests.
  • Serum blanks 4 per 100 serum samples processed may also be diluted 1/10 with 0.1% TFA (25 ⁇ l), combined with the digestion buffer without the incorporation of trypsin before, and after, ZipTip cleanup and incubated overnight at 37°C.
  • a method of identifying a subject having a dementia brain disease, for example AD or MCI, or following the progression of such a brain disease comprising:
  • SPARC-like 1 precursor fibrinogen alpha chain precursor, contactin-1 precursor and serum albumin precursor
  • nucleic acid probes or primers capable of specifically hybridising to, for example mRNA encoding the proteins or peptides may be used.
  • nucleic acid molecule can hybridise to nucleic acid molecules according to the invention under conditions of high stringency.
  • Typical conditions for high stringency include 0.1 x SET, 0.1% SDS at 68 0 C for 20 minutes.
  • Quantitative techniques for the detection of gene expression products are generally known in the art and include real time PCR. This is a generally well-known technique that typically uses fluorescent dyes that bind to double-stranded DNA or DNA oligoprobes that fluoresce when hybridised to a complementary DNA.
  • Proteins, peptides or fragments may be identified or quantified by a variety of techniques generally known in the art. For example, antibodies specific to the proteins, peptides or fragments may be produced by the well-known Kohler & Milstein method. These may be used in assays, such as ELISA assays. The antibodies may be labelled, directly or indirectly, with a dye or radioisotope to allow detection and quantifying of antibody bound to the protein, peptide or fragment.
  • fragment we mean a fragment of the protein or peptide capable of being specifically bound by an antibody raised against the protein or peptide and that allows the identification of that fragment as a frequent of the protein or peptide.
  • antibody includes intact, molecules and fragments such as Fa, F(ab') 2 and Fv.
  • the methods of the invention may additionally comprise determining the level in a test subject of tau protein and/or amyloid beta (AB) or a fragment thereof.
  • the AJi may be the 1- 42 fragment of A ⁇ .
  • a further aspect of the invention provides a kit for use in a method of detecting or monitoring a dementia brain disorder, the kit comprising positive controls for mass spectrometry analysis in which the kit comprises:
  • the kit comprises a control sample representative of a subject not having the disorder comprises proteins and/or proteolytically digested peptides which when analysed by mass spectrometry generate two or more ions as described in the first aspect of the invention.
  • a still further aspect of the invention relates to a method of identifying a drug for treating a dementia brain disorder:
  • a compound which is suitable for treating dementia would cause the peaks of the fingerprint to match more closely to a fingerprint from a control representing a non-disorder sample, than to a fingerprint from a control representing a dementia sample.
  • a further aspect of the invention relates to a computer system having a processor, a memory and an input for receiving data from a mass spectrometer, wherein the memory comprises data corresponding to a fingerprint according to the first aspect of the invention and the computer configured to compare data from the input against the fingerprint data and to identify whether the inputted data is indicative of a dementia brain disorder based on the results of the comparison.
  • the invention also provides a disk, for example CD or diskette, on which one or more of the fingerprints of the first aspect of the invention is stored. Furthermore, the invention provides one or more fingerprints accessible via the internet.
  • the invention also provides an assay kit for the detection of the presence or absence of a brain disorder or for the progression of a dementia brain disorder comprising one or more detection elements for specifically identifying a level of one or more of the following proteins, peptides or fragments thereof in a sample selected from:
  • SPARC-like 1 precursor fibrinogen alpha chain precursor, contactin-1 precursor and serum albumin precursor.
  • 2 or 3 of the proteins or peptides may have detection elements.
  • the assay kit comprises detection elements specific for each of the following: SPARC-like 1 precursor, fibrinogen alpha chain precursor, contactin-1 precursor and serum albumin precursor.
  • the assay list may additionally comprise a detection element for tau protein and/or amyloid beta (AB) or a fragment thereof.
  • AB amyloid beta
  • the fragment is the 1-42 fragment of AB.
  • the assay kit preferably has one or more detection elements selected from:
  • nucleic acid molecule capable of specifically hybridising to a nucleic acid molecule encoding the protein, peptide or fragment thereof; or (ii) an antibody capable of specifically binding to the protein, peptide or fragment thereof.
  • antibodies and nucleic acid molecules such as probes or primers, may be made by methods well-known in the art.
  • Figure 2B Magnified peptide spectra over the mass range 1000 to 2500 from the data shown above. All intensities have been equalised to allow comparisons to be made; areas of interest are depicted by the red boxes.
  • Figure 3A Distribution of biomarker ion performances in step 1 of the stepwise analysis.
  • Figure 3B Distribution of the top 100 biomarker ion performances in step 1 of the stepwise analysis.
  • Figure 3C Population distribution based on mean predicted class value over 50 random sample cross validations for the cases used to train the ANN model when considered as unseen cases (to the model). Hashed bars represent Control, solid bars represent AD samples.
  • Figure 3D Response graphs generated singly for each of the 4 ions used in the model indicating the relationship between intensity for an ion of a given m/z value and predicted disease class.
  • Figure 4 Modelling of the data from 54 CSF samples (20 controls and 34 AD) and a validation set of 10 MCI CSF samples.
  • Lumbar puncture was performed in the L3/L4 or L4/L5 interspace with the subject in the sitting position. The first mL of CSF was discarded, 1 mL was sent for cell analysis and 10 mL were collected in plastic (polypropylene) tubes. All CSF samples were gently mixed to avoid possible gradient effects. No CSF sample contained more than 500 erytrocytes/ ⁇ L. The CSF samples were centrifuged at 2000 x g at 4°C for 10 min to eliminate cells and other insoluble material, and were then immediately frozen and stored at -80 0 C pending biochemical analyses.
  • QC Quality Control
  • Trifluroacetic acid (1 OTFA) blanks Trifluroacetic acid (1 OTFA) blanks, 5 QC samples (serum samples clotted for 30 minutes exactly and aliquoted within 1 hour of collection and stored at -80°C), 5 Bovine Serum Albumin controls (peptides only) and 5 sample blanks are included in all protocols for every 50 samples processed. Further to this, the protein and peptide analyses are run for 8 QC serum samples on a weekly basis and the samples subjected to a basic cluster analysis. This is a very important aspect of the QC procedure, because to date if the samples from a single run cluster as a separate population then the sample preparation including reagents and operator are checked as a problem area.
  • Protein and Peptide Fingerprints Preliminary experiments were conducted using various dilutions of the CSF samples with 0.1% TFA (i.e. 1 in 2, 1 in 5 and undiluted CSF).
  • the diluted samples were ZipTip fractionated for two reasons: firstly to remove of high abundant proteins in order to reduce the complexity of the samples which are known to cause suppression of the low molecular weight and secondly to concentrate the proteins/peptides by eluting these in a lower volume of organic solvent.
  • Ci 8 Zip Tips were chosen as these are most suitable for low molecular protein and peptides ⁇ 50,000 Da.
  • Sample Preparation - Protein and Peptide analysis Prior to sample handling and preparation, the samples were randomized, for position on a 96-well dilution plate and a 384 MALDI target plate, using a computer randomization program (Microsoft Excel). All samples were prepared in duplicate. Briefly, neat CSF (25 ⁇ L) was fractionated using a Ci 8 ZipTip , (Milipore, Watford, UK) with 25 cycles of binding. This was followed by two washes in 0.1 % TFA and elution in 8 ⁇ L of 80 % acetonitrile/0.1 % TFA.
  • the reaction was quenched with 1% TFA (l ⁇ l) and the sample cleaned-up using a Ci 8 ZipTip following the procedure described above and an aliquot of the eluate (l ⁇ l) was spotted onto the MALDI target using the dried droplet method with matrix, CHCA (Laser BioLabs, Cedex, France), prepared as a 10 mg/ml solution in 50 % ACN + 0.1 % TFA for tryptic peptide analysis. Duplicate samples were applied to the target plate in the randomized order used for sample preparation and analysed by MALDI/TOF-MS. A range of appropriate blank and control samples will be prepared alongside the serum digests.
  • Mass Spectrometer Calibration The inventors have opted to use close external calibration method for all the experiments carried out.
  • the calibration mix obtained from Laser Biolabs (Cedex, France) consists of three calibrants, but to ensure that we have a calibrant within each of the acquired mass ranges used to capture the mass spectra for the proteins (i.e. 1 - 30,000 Da), we add insulin ⁇ -chain to the mixture.
  • the protein calibrants are: cytochrome C, horse heart, m/z 12361.12; myoglobin, horse, m/z 16181.06 and trypsinogen, m/z 23981.98. Insulin beta chain, m/z 3494.65 (3 ⁇ l of 5mM) was also added to the calibration protein mix.
  • the peptide calibration was based on the monoisotopic masses of [M+H] + of bradykinin fragment 1-5, angiotensin II, neurotensin, ACTH clip (18-39) and Insulin B-chain oxidised at m/z 573.31, 1046.54, 1672.91, 2465.19, 3494.65, respectively.
  • the calibrants were mixed with matrix beforehand to ensure a homogeneous mix. For a full 384 sample spot plate with 96 calibration spots 20 ⁇ l of calibrant was mixed with either 80 ⁇ l SA or CHCA before spotting 0.5 ⁇ l in between the sample spots.
  • Mass Spectrometric Analysis MALDI-TOF experiments were performed on the Axima CFR + mass spectrometer (Shimadzu, Manchester, UK). Mass spectral acquisition in the range 1000 - 30000 Da was carried out in 'raster mode' for proteins using linear TOF-MS whilst tryptic peptides were analysed in reflectron mode in the range 800 - 3500 Da using pre- scanning in 'auto quality' mode. Laser shots were set to 5 shots per profile and a cumulative accumulation 100 profiles were captured using automated Kompact software (Shimadzu, Manchester, U.K.). The automated programme allows the update of calibration for every 4 samples analysed which can then be visually checked for mass accuracy.
  • the six-port valve was switched to introduce a counter-current solvent flow (180 nL/min, split flow) to the pre-column from an UltiMate gradient pump (LC Packings, Dionex Ltd, UK) to direct the sample onto an reverse-phase capillary LC column (LC Packings, Cjg-PepMap, 100 A 5 3 ⁇ m particle size, 75 ⁇ m ID x 150 mm, Dionex Ltd, UK) connected to the mass spectrometer interface by a fused silica transfer line (20 ⁇ m ID, 300 mm).
  • LC Packings, Cjg-PepMap 100 A 5 3 ⁇ m particle size, 75 ⁇ m ID x 150 mm, Dionex Ltd, UK
  • On-line sample separation prior to mass spectrometric detection was carried out using a linear gradient (Solvent A: 0.1 % formic acid in water; Solvent B: 80 % acetonitrile in 0.1 % formic acid in water) from 5 % B (at time of six-port valve switch) to 75 % B (over 60 min., hold for 2 min.) then to 5 % B (over 2 min.) and re-equilibrate for the next analytical run.
  • Tandem mass spectrometric analysis was carried out using a Finnigan LCQ Classic ion trap mass spectrometer (ThermoElectron, San Jose, CA, USA) equipped with a dynamic nano-electrospray ion source, operated in positive ion mode.
  • Analytical performance of the hyphenated LC-QIT MS/MS system was assessed by analysis of a BSA tryptic digest as standard, bracketing replicate serum samples, to provide validation for human serum analysis. Data was acquired for human serum tryptic peptides following LC introduction using either full scan mode (m/z 300-2000, 3 microscans) or targeted tandem mass spectrometric (MS/MS) mode (200 ms activation time, isolation m/z of 3.0). Automatic gain control (AGC) was applied in all data acquisition modes.
  • full scan mode m/z 300-2000, 3 microscans
  • MS/MS targeted tandem mass spectrometric
  • AGC Automatic gain control
  • Sequence identities were confirmed using the Mascot database and the search parameter settings for the MASCOT sequence query routine were as follows: 1 maximum missed cleavage, 0.8 Da tolerance was used for the singly and doubly charged precursor ion and 0.5 Da for the fragment ion mass. NCIBnr was used as the reference database (human taxonomy). Trypsin was set as the proteolytic enzyme.
  • Computational algorithm and architecture The basis of the analysis within this study was a back propagation algorithm applied to a multi-layer perceptron architecture using a supervised learning approach. Learning rate and momentum were set at 0.1 and 0.9 respectively.
  • Learning rate and momentum were set at 0.1 and 0.9 respectively.
  • This architecture was constrained by utilizing only two hidden nodes and randomized initial weights with a small range about zero. This approach was adopted by Ball et al, 3 in an earlier study as a method of amplifying the importance of key ions within highly dimensional systems such as mass spectrometry data, while producing accurate predictions and maintaining generalization.
  • Cross validation Prior to training of each model the data was randomly divided into three sub sets; a training subset, a test subset and a validation subset to allow random sample cross validation. These subsets consisted of splits of 60%, 20 % and 20 % respectively. This process is repeated prior to training for 50 sub-models creating 50 different training data, test data and validation data splits. The validation data set is not used in anyway during training and is therefore used to assess the model performance for blind data. This double validation provides a more realistic representation of model performance than using a single validation data set 21 .
  • Marker identification The modelling process involved a novel stepwise approach essentially a wrapper method 22 with iterative forward selection coupled with marker selection based on performance on unseen data. Initially, each variable from the dataset is be used as an individual input in a network, thus creating n (23001) individual models. These n models are then trained, using the random sample cross validation process described above, creating 50 sub-models for each of the n models. These models are ranked in ascending order based upon their mean squared error values for test data. The model with the lowest predictive error identifies the most important single ion which will be selected for inclusion in the subsequent additive step.
  • n-1 models each containing two inputs. Training is repeated and performance evaluated for these as described above. The model which showed the best performance is then selected and the process repeated, creating n-2 three input models. This process is repeated until no statistically significant improvement in model performance is gained from the addition of further inputs for three complete steps, thus resulting in a final model containing the proteomic pattern which most accurately predicts between the two outcomes.
  • This process will define a best subset of ions (that are ordinal to each other) that explain the variation within the disease versus control classes.
  • Mass spectrometry profiles Technical replicates for control and AD samples were shown to have a mass-to-charge ratio variation of between 0.01 - 0.05% across a mass range of 1000-30000 Da and intensity variation across the same mass range of 15% for a serum proteome (data not presented). These variation rates are well within what is considered acceptable and indicate a highly reproducible profile as generated by the mass spectrometry instrument.
  • the CSF samples were used to successfully generate mass spectrometry profiles of good reproducibility examples of which are shown in Figure 1 and 2A and 2B.
  • Stepwise model development and optimisation In step 1 performance on blind data ranged from 43.2% correct classification to 83.8% correct classification depending on the single ion used within the model. The most important single ion had a mass value of 13645 Da. These performances followed a clear normal distribution (Figure 3A.). The average performance across all of the ions was 63.7 percent correct classification. When a close examination of the top 100 ions was conducted clear steps in performance were observed ( Figure 3B.).
  • Stepwise analysis The addition of further ions through the stepwise approach (Table 1) improved the performance for test data in subsequent steps even though no large improvement was seen in the selection performance (the basis for inclusion of the ion) until the fourth step. In this case the information in subsequent ions is explaining only one or 2 additional cases.
  • Characterisation of marker ions Response curves were generated for each of the ions in the 4 ion model by presenting the model with varying values within the range of those found in the data and maintaining the other ions as the mean value. In this way the responses of ions were investigated singly for m/z values of 13014, 15344, 13029 and 12952 respectively ( Figures 3D.).
  • ion 1 (m/z 13014) had a strong positive influence at lower intensities and a weaker positive influence at higher intensities with respect to the presence of AD; ion 2 (m/z 15344) had a weak positive influence; ion 3 (m/z 13029) had a weak positive influence and ion 4 (m/z 12952) had a strong negative influence.
  • ANNs identified biomarker patterns containing seven ions from the tryptically digested peptide profiles, which correctly discriminated between control and AD samples to a median accuracy of 92.73% (inter-quartile range 89.4 - 94.8 %, sensitivity of 85 % and specificity of 97.14 %.
  • Table 2. shows the performance of the model at each step of the tryptic peptide data.
  • biomarker ions within these models were considered to be those which most accurately distinguished between AD patients and healthy control CSF for this dataset. Averaging across a suite of 50 models achieves with an AUC of 0.9935. When this model was applied to data from MCI patients all but case were predicted correctly. Further work is needed to resolve this issue and possibly refine our current model to improve its predictive power.
  • MALDI mass spectrometry coupled with computational techniques offers a clear and significant potential for the identification and determination of the characteristics of AD.
  • Clearly preliminary results presented herein indicate it is possible to potentially develop diagnostic models of disease that have high sensitivity and specificity for the diagnosis of AD for unseen cases. These models can utilize a greatly reduced panel of biomarkers rather than relying on a complete spectral profile.
  • Biomarker ions A proteomic approach allows parallel analysis of hundreds of proteins in one single experiment. Current advances in the proteomic field have been applied to screen for new biomarkers specific for AD. A number of protein differences in particular increased levels of tau protein and reduced levels of 1—42 fragment of amyloid precursor protein have been reported in CSF of AD patients 26 . However, there still remains to be found a sensitive and specific biomarker analysis for AD. Several studies using 2-D gel and SELDI of CSF proteins in AD patients have been reported 27"36 . The main limitation of this methodology as already mentioned is that large samples sets cannot be analysed in a reproducible manner. Therefore a study of CSF proteins has also been carried out on nine AD patients compared to ten healthy controls analysed by SELDI-MS 32 .
  • This panel of biomarkers has not previously been associated with AD pathology moreover, the specificity of this panel can be gleamed from the ANNs model when MCI patients were used as a totally blind set and it not only predicted these correctly as non-control samples but even more remarkable it was able to predict these between the control and AD patients.
  • SPARCLl has antiadhesive properties [2,4], and loss of SPARCLl expression, possibly due to deletions/mutations or promoter hypermethylation [26], is associated with increased proliferative activity and cell cycle progression [24].
  • the expression of SPARC is limited largely to tissues undergoing repair or remodeling due to wound healing, disease, or natural processes. Pathologies such as cancer metastasis, arthritis, diabetes, and kidney disease are characterized by elevated expression of SPARCZReed and Sage, 1996. Elucidating the mechanism by which SPARC contributes to these pathologies will lead to a better understanding of the biology of this important matrix- associated protein and the function of matricellular proteins in human disease.
  • Fibrinogen alpha has been reported to be implicated in hereditary systemic amyloidosis which is caused by deposition of genetically variant proteins as amyloid fibrils.
  • the types that present with renal disease are usually associated with mutations in the genes for apolipoprotein AI, apolipoprotein All, lysozyme or fibrinogen A a-chain. It has been postulated that a similar type of deposition due to dysfunctional Fibrinogen alpha protein may be present in Ad patients.
  • ANNs have previously been considered by others a black box and have received criticism that they provide little insight into the system being modelled.
  • the inventors present methodologies that address this criticism by fully interrogating the models developed to gain significant information about the population, the data and the biological system.
  • ANNs have the potential to suffer from "the curse of dimensionality" when being utilised for biomarker identification from highly dimensional data sets.
  • biomarkers that are not only highly specific and sensitive for detection of given clinical conditions but also have biological relevance and are not likely to be epiphenomena. Thus it is likely the factors identified are relevant to the biology of the system being investigated.
  • Model performance Clearly a large number of cases are explained by a single biomarker ion 83.8 % explained by ion 13645. The performance of the top ion is broadly comparable to those achieved by current cognitive methods for late stage disease. Examination of the predictions of the model indicated a clear structure in the population could be defined. In this way potential outliers could be identified. These cases would warrant further investigation as they have an indicated risk profile that is not in concordance with the rest of the population. It is conceivable that these cases are individuals misclassified by cognitive methods and therefore there appears to be a clear limitation within the AD dataset given that the assessment of these patients is based on a diagnostic method which has sources of error and relies on the experience of clinical personnel and studies have shown that the accuracy of the clinical diagnosis is between 65 and 90%.
  • AD population within this study have been classified with cognitive methods which we believe to be an important aspect in our study which has enabled us to identify of a panel of biomarkers consistent with AD from the CSF proteome/peptidome which can be applied to the whole study population.
  • cognitive methods which we believe to be an important aspect in our study which has enabled us to identify of a panel of biomarkers consistent with AD from the CSF proteome/peptidome which can be applied to the whole study population.

Abstract

L'invention concerne une empreinte de l'état d'un cerveau atteint de démence, obtenue par spectroscopie de masse, pour identifier dans le cerveau la présence d'un état de démence et/ou pour identifier dans le cerveau la progression d'un état de démence, et comprenant les pics dérivés des ions suivants : les ions provenant d'un peptide digéré par protéolyse, les ions ayant des valeurs (m/z): 2048, 1434 et 1336 (+/- 0,02 %) et éventuellement un ou plusieurs des ions supplémentaires suivants, les ions supplémentaires ayant des valeurs (m/z): 2734, 1018, 3225 et/ou 3121 (+/- 0,02 %). L'invention propose également des procédés permettant d'identifier des troubles de démence dans le cerveau.
PCT/GB2009/001624 2008-06-27 2009-06-29 Détermination WO2009156747A2 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
GB0811858A GB0811858D0 (en) 2008-06-27 2008-06-27 Assay
GB0811858.0 2008-06-27

Publications (2)

Publication Number Publication Date
WO2009156747A2 true WO2009156747A2 (fr) 2009-12-30
WO2009156747A3 WO2009156747A3 (fr) 2010-02-25

Family

ID=39683316

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/GB2009/001624 WO2009156747A2 (fr) 2008-06-27 2009-06-29 Détermination

Country Status (2)

Country Link
GB (1) GB0811858D0 (fr)
WO (1) WO2009156747A2 (fr)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012142300A2 (fr) * 2011-04-12 2012-10-18 Quanterix Corporation Procédé de détermination d'un protocole de traitement et/ou d'un pronostic de rétablissement d'un patient à la suite d'un traumatisme cérébral provenant d'un événement hypoxique
WO2012142301A2 (fr) * 2011-04-12 2012-10-18 Quanterix Corporation Procédé de détermination d'un protocole de traitement et/ou d'un pronostic de rétablissement d'un patient à la suite d'un traumatisme cérébral
WO2018009727A1 (fr) * 2016-07-06 2018-01-11 Brigham Young University Procédés permettant de diagnostiquer et de traiter la maladie d'alzheimer
CN107765015A (zh) * 2017-09-08 2018-03-06 中国计量科学研究院 一种定量检测脑组织中Aβ40蛋白的方法

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
ABDI FADI ET AL: "Detection of biomarkers with a multiplex quantitative proteomic platform in cerebrospinal fluid of patients with neurodegenerative disorders" JOURNAL OF ALZHEIMER'S DISEASE, IOS PRESS, AMSTERDAM, NL, vol. 9, no. 3, 1 August 2006 (2006-08-01), pages 293-348, XP009090146 ISSN: 1387-2877 *
Ball G. et al.: "Diagnosis of Alzheimer's disease from cerebro-spinal fluid"[Online] 24 July 2007 (2007-07-24), XP002550115 Retrieved from the Internet: URL:http://www.alzheimers-research.org.uk/assets/docs/20070718143555CIMED_programme.pdf> [retrieved on 2009-10-13] *
Matharoo-Ball B. et al.: Alzheimer's and Dementia, vol. 2, no. 3 Suppl 1, 2007, page S283, XP002550127 DOI: 10.1016/j.jalz.2006.05.1005 *
Matharoo-Ball: "Computational Methods for the diagnosis of Alzheimer's disease from cerebro-spinal fluid by matrix assisted laser desorption-ionisation mass spectrometry (MALDI-MS)"[Online] 10 July 2007 (2007-07-10), XP002550116 Retrieved from the Internet: URL:http://www.tech.plym.ac.uk/spmc/pdf/CIMED2007/S3P7.pdf> [retrieved on 2009-10-13] *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012142300A2 (fr) * 2011-04-12 2012-10-18 Quanterix Corporation Procédé de détermination d'un protocole de traitement et/ou d'un pronostic de rétablissement d'un patient à la suite d'un traumatisme cérébral provenant d'un événement hypoxique
WO2012142301A2 (fr) * 2011-04-12 2012-10-18 Quanterix Corporation Procédé de détermination d'un protocole de traitement et/ou d'un pronostic de rétablissement d'un patient à la suite d'un traumatisme cérébral
WO2012142301A3 (fr) * 2011-04-12 2013-01-10 Quanterix Corporation Procédé de détermination d'un protocole de traitement et/ou d'un pronostic de rétablissement d'un patient à la suite d'un traumatisme cérébral
WO2012142300A3 (fr) * 2011-04-12 2013-01-31 Quanterix Corporation Procédé de détermination d'un protocole de traitement et/ou d'un pronostic de rétablissement d'un patient à la suite d'un traumatisme cérébral provenant d'un événement hypoxique
US10393759B2 (en) 2011-04-12 2019-08-27 Quanterix Corporation Methods of determining a treatment protocol for and/or a prognosis of a patient's recovery from a brain injury
US11275092B2 (en) 2011-04-12 2022-03-15 Quanterix Corporation Methods of determining a treatment protocol for and/or a prognosis of a patient's recovery from a brain injury
WO2018009727A1 (fr) * 2016-07-06 2018-01-11 Brigham Young University Procédés permettant de diagnostiquer et de traiter la maladie d'alzheimer
US11719708B2 (en) 2016-07-06 2023-08-08 Brigham Young University Diagnosing and treating alzheimer's disease
CN107765015A (zh) * 2017-09-08 2018-03-06 中国计量科学研究院 一种定量检测脑组织中Aβ40蛋白的方法

Also Published As

Publication number Publication date
GB0811858D0 (en) 2008-07-30
WO2009156747A3 (fr) 2010-02-25

Similar Documents

Publication Publication Date Title
Fusaro et al. Prediction of high-responding peptides for targeted protein assays by mass spectrometry
Zhang et al. Evaluation of a novel, integrated approach using functionalized magnetic beads, bench-top MALDI-TOF-MS with prestructured sample supports, and pattern recognition software for profiling potential biomarkers in human plasma
Seibert et al. Advances in clinical cancer proteomics: SELDI-ToF-mass spectrometry and biomarker discovery
Veenstra Global and targeted quantitative proteomics for biomarker discovery
Lopez et al. High-resolution serum proteomic profiling of Alzheimer disease samples reveals disease-specific, carrier-protein–bound mass signatures
US8389222B2 (en) Apolipoprotein fingerprinting technique and methods related thereto
Schilde et al. Protein variability in cerebrospinal fluid and its possible implications for neurological protein biomarker research
Wu et al. Identification of a uromodulin fragment for diagnosis of IgA nephropathy
US20080086272A1 (en) Identification and use of biomarkers for the diagnosis and the prognosis of inflammatory diseases
JP2008545960A (ja) 結核の診断
BRPI0711967A2 (pt) ensaio de biomarcador de espectrometria de massa
Gianazza et al. The selected reaction monitoring/multiple reaction monitoring-based mass spectrometry approach for the accurate quantitation of proteins: clinical applications in the cardiovascular diseases
US20160123997A1 (en) Materials and methods relating to alzheimer's disease
Filip et al. Advances in urinary proteome analysis and applications in systems biology
Watson et al. Quantitative mass spectrometry analysis of cerebrospinal fluid protein biomarkers in Alzheimer’s Disease
Niu et al. Specific serum protein biomarkers of rheumatoid arthritis detected by MALDI-TOF-MS combined with magnetic beads
WO2009156747A2 (fr) Détermination
Velstra et al. Improved classification of breast cancer peptide and protein profiles by combining two serum workup procedures
CN112305122B (zh) 代谢物标志物及其在疾病中的应用
Callesen et al. Reproducibility of serum protein profiling by systematic assessment using solid‐phase extraction and matrix‐assisted laser desorption/ionization mass spectrometry
O'riordan et al. Technology Insight: renal proteomics—at the crossroads between promise and problems
Yang et al. Serum peptidome profiling in patients with lung cancer
Berven et al. Proteomic-based biomarker discovery with emphasis on cerebrospinal fluid and multiple sclerosis
Liu et al. Proteomic profiling of occupational medicamentosa-like dermatitis induced by trichloroethylene in serum based on MALDI-TOF MS
Dekker et al. Peptide profiling of cerebrospinal fluid by mass spectrometry

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 09769588

Country of ref document: EP

Kind code of ref document: A2

NENP Non-entry into the national phase in:

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 09769588

Country of ref document: EP

Kind code of ref document: A2