+ All Categories
Home > Documents > O-PLS & Metabonomics

O-PLS & Metabonomics

Date post: 10-Apr-2018
Category:
Upload: parvanehe
View: 225 times
Download: 0 times
Share this document with a friend

of 24

Transcript
  • 8/8/2019 O-PLS & Metabonomics

    1/24

    By: Parvaneh EbrahimiBy: Parvaneh Ebrahimi

    JulyJuly--20102010

  • 8/8/2019 O-PLS & Metabonomics

    2/24

    Recently, postgenomic technologies, including transcriptomic andproteomic methods, have been developing as a means of generating

    high information content biological assays for disease diagnoses or for

    evaluating the beneficial or adverse effects of pharmaceuticals

    INTRODUCTIONINTRODUCTION

  • 8/8/2019 O-PLS & Metabonomics

    3/24

    MetabonomicsMetabonomics is an emerging complementary postgenomic technology

    and involves the determination of the levels and time dependence of

    low molecular mass endogenous metabolites in biofluids such as

    urine, plasma, or in tissues, as a result of some pathophysiological

    effect.

    Most studies have used 1H nuclear magnetic resonance spectroscopy

    (1H NMR) for data generation, but more recently, liquid

    chromatography hyphenated with mass spectrometry, LC/MS, has

    been employed.

    In other types of biological systems, GC/MS has been used extensively.

  • 8/8/2019 O-PLS & Metabonomics

    4/24

    Megavariate pattern recognition methods are used to extract relevantbiological information from the complex data.

    Unsupervised or supervised chemometrics methods, such as principal

    component analysis (PCA) or discriminant analysis by projection

    on latent structures (PLS-DA), are often used.

    In HNMR, amount of the data is huge and peaks corresponding to

    certain molecules, such as citrate, can be subject to peak positionvariation from sample to sample. The chief cause of such

    variability is pH differences between samples.

    .

  • 8/8/2019 O-PLS & Metabonomics

    5/24

    Methods that reduce the amount of data down to manageableproportions have been used.

    One of the most common methods involves signal integration within

    spectral regions orb

    ins , and these have generally correspondedto a typical spectral width of 0.04 ppm.

    This data reduction method also has the property of largely

    concealing any peak position variation and thus providing more

    robust modeling.

  • 8/8/2019 O-PLS & Metabonomics

    6/24

  • 8/8/2019 O-PLS & Metabonomics

    7/24

    Pattern RecognitionPattern Recognition

    The multivariate pattern recognition method used in this paper is

    O-PLS, written as:

    StructuredNoise

  • 8/8/2019 O-PLS & Metabonomics

    8/24

    For the applications presented here, each line of the X matrix is an

    NMR spectrum corresponding to one sample and each column

    of Y defines a class (or group) whose values are dummy variables.

    The method can therefore be defined as O-PLS-DA.

    The O-PLS loading and regression coefficients allow for a morerealistic interpretation than PLS, because in the O-PLS,

    the structured noise is modeled separately from the variation

    common toX and Y .

  • 8/8/2019 O-PLS & Metabonomics

    9/24

    To test the validity of models against over-fitting, the cross-validation

    parameter was computed:

  • 8/8/2019 O-PLS & Metabonomics

    10/24

    EXPERIMENTAL SECTIONEXPERIMENTAL SECTION

    To 62 male Sprague-Dawley rats ,saline(0.9% ) alone or with

    was administrated.1-mg.kg75.0at a dose level of2HgCl

    Urine samples were collected continuously at various time intervals:

    (predose and 0-8, 8-24, 24-32, 32-48, 48-72, 72-96, 96-120, 120-144,and 144-168 hr after treatment).

    To minimize any gross variation in the pH of the urine samples,

    Phosphate buffer solution was used to fix the samples PH in (7.4).

    Urine samples were centrifuged in order to remove particulatecontaminants, and the samples were stored at -40 C pending NMRspectroscopic analysis.

  • 8/8/2019 O-PLS & Metabonomics

    11/24

    For each sample, an 1H NMR spectrum was measured at 600.13MHz

    on a BrukerDRX-600 spectrometer.

    To make a comparison of the use of real NMR spectra and reduced

    spectra, the original binning methodology was also applied, andhere each NMR spectrum was segmented into 250 chemical shift

    regions of 0.04 ppm.

    Finally, to take account of large variations in urine concentration, using

    both reduced and full resolution data sets, all spectra were then

    normalized.

  • 8/8/2019 O-PLS & Metabonomics

    12/24

    After data acquisition, the spectra of the 62 rat urine samples were

    separated into two groups corresponding to the control and treatedgroups and finally an O-PLS model is generated.

  • 8/8/2019 O-PLS & Metabonomics

    13/24

    Citrate is known to be one of the compounds that have the largest peak

    position variation, mainly due to pH and metal ion variation.

    It is obvious that the impact of this peak position variation on pattern

    recognition models must be important.

  • 8/8/2019 O-PLS & Metabonomics

    14/24

    SIMULATIONSIMULATIONA simulation study based on the variation of only three compounds that

    are subject to peak position variation may provide the information

    necessary for accurate interpretation of more complex data sets such

    as in a full toxicological study.

    A biofluid NM

    R spectrum can be considered as a linear combination ofspectra of pure single molecule components.

  • 8/8/2019 O-PLS & Metabonomics

    15/24

    several molecules were chosen as chemical compounds of interest

    because of their known peak position variation or overlap. These were

    citrate, taurine, and trimethylamine N-oxide (TMAO).The last two molecules havebeen chosen not only because their

    respective signals are subject to positional noise but also because they

    overlap, a common feature of real NMR spectra.

  • 8/8/2019 O-PLS & Metabonomics

    16/24

  • 8/8/2019 O-PLS & Metabonomics

    17/24

    without peak position variationwithout peak position variation

  • 8/8/2019 O-PLS & Metabonomics

    18/24

  • 8/8/2019 O-PLS & Metabonomics

    19/24

    molecule prone to a peak position variation but which is notmolecule prone to a peak position variation but which is not

    related to the discrimination between the two groupsrelated to the discrimination between the two groups

  • 8/8/2019 O-PLS & Metabonomics

    20/24

  • 8/8/2019 O-PLS & Metabonomics

    21/24

    variation which is related to thevariation which is related to the

    discrimination between the twodiscrimination between the twogroupsgroups

  • 8/8/2019 O-PLS & Metabonomics

    22/24

    CONCLUSIONCONCLUSION

    In this paper, it was demonstrated that metabonomic studies can be

    carried out without any reduction of the true 1H NMR spectral

    resolution.

    From simulation studies, variation of peak position can reduce slightly

    the prediction ability of the model, but at the same time , this peak

    position variation can be modeled by the pattern recognition

    method and therefore may provide:

    useful information about physicochemical variations in the

    biofluid Matrix.

  • 8/8/2019 O-PLS & Metabonomics

    23/24

    CONCLUSIONCONCLUSION

    For any other analytical techniques (i.e., LC-MS), where peak position

    variation is often considered as problematic, it may also be the case.

    The complexity of the loading interpretation due to the high number of

    variables can be simplified using O-PLS, which reduces the number

    of loadings to be interpreted.

  • 8/8/2019 O-PLS & Metabonomics

    24/24

    THANK YOUTHANK YOU


Recommended