Prediction of Cell Wall Properties and Response to Deconstruction Using Alkaline Pretreatment in Diverse Maize Genotypes Using Py-MBMS and NIR

Muyang Li, Daniel L. Williams, Marlies Heckwolf, Natalia de Leon, Shawn Kaeppler, Robert W. Sykes, David Hodge

    Research output: Contribution to journalArticle

    Abstract

    In this work, we explore the ability of several characterization approaches for phenotyping to extract information about plant cell wall properties in diverse maize genotypes with the goal of identifying approaches that could be used to predict the plant’s response to deconstruction in a biomass-to-biofuel process. Specifically, a maize diversity panel was subjected to two high-throughput biomass characterization approaches, pyrolysis molecular beam mass spectrometry (py-MBMS) and near-infrared (NIR) spectroscopy, and chemometric models to predict a number of plant cell wall properties as well as enzymatic hydrolysis yields of glucose following either no pretreatment or with mild alkaline pretreatment. These were compared to multiple linear regression (MLR) models developed from quantified properties. We were able to demonstrate that direct correlations to specific mass spectrometry ions from pyrolysis as well as characteristic regions of the second derivative of the NIR spectrum regions were comparable in their predictive capability to partial least squares (PLS) models for p-coumarate content, while the direct correlation to the spectral data was superior to the PLS for Klason lignin content and guaiacyl monomer release by thioacidolysis as assessed by cross-validation. The PLS models for prediction of hydrolysis yields using either py-MBMS or NIR spectra were superior to MLR models based on quantified properties for unpretreated biomass. However, the PLS models using the two high-throughput characterization approaches could not predict hydrolysis following alkaline pretreatment while MLR models based on quantified properties could. This is likely a consequence of quantified properties including some assessments of pretreated biomass, while the py-MBMS and NIR only utilized untreated biomass.

    Original languageEnglish (US)
    Pages (from-to)1-15
    Number of pages15
    JournalBioenergy Research
    DOIs
    StateAccepted/In press - Oct 4 2016

    Profile

    biomass
    Biomass
    pyrolysis
    least squares
    pretreatment
    mass spectrometry
    Mass spectrometry
    Pyrolysis
    Infrared radiation
    linear models
    cell walls
    corn
    Molecular beams
    Linear regression
    prediction
    genotype
    Hydrolysis
    Throughput
    thioacidolysis
    chemometrics

    Keywords

    • Analytical pyrolysis
    • Chemometrics
    • NIR
    • Plant cell wall property prediction
    • PLS regression
    • Py-MBMS

    ASJC Scopus subject areas

    • Renewable Energy, Sustainability and the Environment
    • Agronomy and Crop Science
    • Energy (miscellaneous)

    Cite this

    Prediction of Cell Wall Properties and Response to Deconstruction Using Alkaline Pretreatment in Diverse Maize Genotypes Using Py-MBMS and NIR. / Li, Muyang; Williams, Daniel L.; Heckwolf, Marlies; de Leon, Natalia; Kaeppler, Shawn; Sykes, Robert W.; Hodge, David.

    In: Bioenergy Research, 04.10.2016, p. 1-15.

    Research output: Contribution to journalArticle

    Li, Muyang; Williams, Daniel L.; Heckwolf, Marlies; de Leon, Natalia; Kaeppler, Shawn; Sykes, Robert W.; Hodge, David / Prediction of Cell Wall Properties and Response to Deconstruction Using Alkaline Pretreatment in Diverse Maize Genotypes Using Py-MBMS and NIR.

    In: Bioenergy Research, 04.10.2016, p. 1-15.

    Research output: Contribution to journalArticle

    @article{9bfff2f7617d45a3b4840029d0f1c03c,
    title = "Prediction of Cell Wall Properties and Response to Deconstruction Using Alkaline Pretreatment in Diverse Maize Genotypes Using Py-MBMS and NIR",
    abstract = "In this work, we explore the ability of several characterization approaches for phenotyping to extract information about plant cell wall properties in diverse maize genotypes with the goal of identifying approaches that could be used to predict the plant’s response to deconstruction in a biomass-to-biofuel process. Specifically, a maize diversity panel was subjected to two high-throughput biomass characterization approaches, pyrolysis molecular beam mass spectrometry (py-MBMS) and near-infrared (NIR) spectroscopy, and chemometric models to predict a number of plant cell wall properties as well as enzymatic hydrolysis yields of glucose following either no pretreatment or with mild alkaline pretreatment. These were compared to multiple linear regression (MLR) models developed from quantified properties. We were able to demonstrate that direct correlations to specific mass spectrometry ions from pyrolysis as well as characteristic regions of the second derivative of the NIR spectrum regions were comparable in their predictive capability to partial least squares (PLS) models for p-coumarate content, while the direct correlation to the spectral data was superior to the PLS for Klason lignin content and guaiacyl monomer release by thioacidolysis as assessed by cross-validation. The PLS models for prediction of hydrolysis yields using either py-MBMS or NIR spectra were superior to MLR models based on quantified properties for unpretreated biomass. However, the PLS models using the two high-throughput characterization approaches could not predict hydrolysis following alkaline pretreatment while MLR models based on quantified properties could. This is likely a consequence of quantified properties including some assessments of pretreated biomass, while the py-MBMS and NIR only utilized untreated biomass.",
    keywords = "Analytical pyrolysis, Chemometrics, NIR, Plant cell wall property prediction, PLS regression, Py-MBMS",
    author = "Muyang Li and Williams, {Daniel L.} and Marlies Heckwolf and {de Leon}, Natalia and Shawn Kaeppler and Sykes, {Robert W.} and David Hodge",
    year = "2016",
    month = "10",
    doi = "10.1007/s12155-016-9798-z",
    pages = "1--15",
    journal = "Bioenergy Research",
    issn = "1939-1234",
    publisher = "Springer New York",

    }

    TY - JOUR

    T1 - Prediction of Cell Wall Properties and Response to Deconstruction Using Alkaline Pretreatment in Diverse Maize Genotypes Using Py-MBMS and NIR

    AU - Li,Muyang

    AU - Williams,Daniel L.

    AU - Heckwolf,Marlies

    AU - de Leon,Natalia

    AU - Kaeppler,Shawn

    AU - Sykes,Robert W.

    AU - Hodge,David

    PY - 2016/10/4

    Y1 - 2016/10/4

    N2 - In this work, we explore the ability of several characterization approaches for phenotyping to extract information about plant cell wall properties in diverse maize genotypes with the goal of identifying approaches that could be used to predict the plant’s response to deconstruction in a biomass-to-biofuel process. Specifically, a maize diversity panel was subjected to two high-throughput biomass characterization approaches, pyrolysis molecular beam mass spectrometry (py-MBMS) and near-infrared (NIR) spectroscopy, and chemometric models to predict a number of plant cell wall properties as well as enzymatic hydrolysis yields of glucose following either no pretreatment or with mild alkaline pretreatment. These were compared to multiple linear regression (MLR) models developed from quantified properties. We were able to demonstrate that direct correlations to specific mass spectrometry ions from pyrolysis as well as characteristic regions of the second derivative of the NIR spectrum regions were comparable in their predictive capability to partial least squares (PLS) models for p-coumarate content, while the direct correlation to the spectral data was superior to the PLS for Klason lignin content and guaiacyl monomer release by thioacidolysis as assessed by cross-validation. The PLS models for prediction of hydrolysis yields using either py-MBMS or NIR spectra were superior to MLR models based on quantified properties for unpretreated biomass. However, the PLS models using the two high-throughput characterization approaches could not predict hydrolysis following alkaline pretreatment while MLR models based on quantified properties could. This is likely a consequence of quantified properties including some assessments of pretreated biomass, while the py-MBMS and NIR only utilized untreated biomass.

    AB - In this work, we explore the ability of several characterization approaches for phenotyping to extract information about plant cell wall properties in diverse maize genotypes with the goal of identifying approaches that could be used to predict the plant’s response to deconstruction in a biomass-to-biofuel process. Specifically, a maize diversity panel was subjected to two high-throughput biomass characterization approaches, pyrolysis molecular beam mass spectrometry (py-MBMS) and near-infrared (NIR) spectroscopy, and chemometric models to predict a number of plant cell wall properties as well as enzymatic hydrolysis yields of glucose following either no pretreatment or with mild alkaline pretreatment. These were compared to multiple linear regression (MLR) models developed from quantified properties. We were able to demonstrate that direct correlations to specific mass spectrometry ions from pyrolysis as well as characteristic regions of the second derivative of the NIR spectrum regions were comparable in their predictive capability to partial least squares (PLS) models for p-coumarate content, while the direct correlation to the spectral data was superior to the PLS for Klason lignin content and guaiacyl monomer release by thioacidolysis as assessed by cross-validation. The PLS models for prediction of hydrolysis yields using either py-MBMS or NIR spectra were superior to MLR models based on quantified properties for unpretreated biomass. However, the PLS models using the two high-throughput characterization approaches could not predict hydrolysis following alkaline pretreatment while MLR models based on quantified properties could. This is likely a consequence of quantified properties including some assessments of pretreated biomass, while the py-MBMS and NIR only utilized untreated biomass.

    KW - Analytical pyrolysis

    KW - Chemometrics

    KW - NIR

    KW - Plant cell wall property prediction

    KW - PLS regression

    KW - Py-MBMS

    UR - http://www.scopus.com/inward/record.url?scp=84989948850&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84989948850&partnerID=8YFLogxK

    U2 - 10.1007/s12155-016-9798-z

    DO - 10.1007/s12155-016-9798-z

    M3 - Article

    SP - 1

    EP - 15

    JO - Bioenergy Research

    T2 - Bioenergy Research

    JF - Bioenergy Research

    SN - 1939-1234

    ER -