Pretreatment of Mass Spectral Profiles:  Application to Proteomic Data

Abstract

Mass spectral profiles are influenced by several factors that have no relation to compositional differences between samples: baseline effects, shifts in mass-to-charge ratio (m/z) (synchronization/alignment problem), structured noise (heteroscedasticity), and, differences in signal intensities (normalization problem). Different procedures for pretreatment of whole mass spectral profiles described by almost 50,000 m/z values are investigated in order to find optimal approaches with respect to revealing the information content in the data. In order to quantitatively assess the impact of different procedures for pretreatment of mass spectral profiles, we use factorial designs with the ratio between intergroup and intragroup (replicate) variance as response. We have examined the influence of smoothing, binning, alignment/synchronization, noise pattern, and normalization on data interpretation. Our analysis shows that the spectral profiles have to be corrected for heteroscedastic noise prior to normalization. An nth root transform, where n is a small, positive integer, is used to create a homoscedastic noise structure without destroying the linear correlation structures describing individual components when using whole mass spectral profiles. The choice of n is decided by a simple graphic procedure using replicate information. Log transform is shown to change the heteroscedastic noise structure from being dominant in high-intensity regions, to produce the largest noise in the low-intensity regions. In addition, log transform has a negative effect on the collinearity in the profiles. Factorial designs reveal strong interactions between several of the pretreatment steps, e.g., noise structure and normalization. This underlines the limited usability of looking at the different pretreatment steps in isolation. Binning turns out to be able to substitute smoothing of spectra by, for example, moving average or Savitsky-Golay, while, at the same time, reducing the data point description of the profiles by 1 order of magnitude. Thus, if the sampling density is high, binning seems to be an attractive option for data reduction without the risk of losing information accompanying the integration of profiles into peaks. In the absence of smoothing, binning should be executed prior to alignment. If binning is not performed, the order of pretreatment should be smoothing, alignment, nth root transform, and normalization.

Opportunities and Limitations for Untargeted Mass Spectrometry Metabolomics to Identify Biologically Active Constituents in Complex Natural Product Mixtures

“…As the nontransformed data sets were able to identify magnolol as active, the false positives associated with magnolol also increased. These results are counter to the findings of other studies . For example, while Arneberg et al found that the n th root transformation positively impacted their models, models built using this transformation were unable to identify active constituents in the current study.…”

“…These results are counter to the findings of other studies . For example, while Arneberg et al found that the n th root transformation positively impacted their models, models built using this transformation were unable to identify active constituents in the current study. These differences may be due to the differences in applications between these two projects.…”

“…It is common practice in metabolomics studies, particularly those utilizing mass spectrometric data, to subject data to a transformation procedure. , Since mass spectrometers are so sensitive in their ability to detect compounds at a wide range of concentrations, they are subject to errors caused by heteroscedastic noise in count data, in which error is proportional to the peak area. , As such, data transformation processes aimed to reduce the error associated with large peak areas are commonly employed. , Many metabolomics projects utilize, for example, a fourth-root transformation of variable peak areas to minimize the impact of heteroscedastic noise and reduce bias towards highly abundant or ionizable compounds. ,, Despite the popularity of this approach, the statistical analysis revealed that this transformation negatively impacted the ability of models to accurately predict active compounds. Models built using transformed data (Figure E–H) gave berberine and magnolol worse selectivity ratio rankings than data sets using nontransformed data (Figure A–D).…”

“…These differences may be due to the differences in applications between these two projects. While Arneberg et al were assessing proteomics data sets, the data sets assessed for this project were focused on metabolomics-driven natural products discovery. In natural products discovery projects, low-abundant constituents that contribute to bioactivity may be present in the upper parts per million or parts per thousand range, while protein biomarkers are often found in the lower parts per billion range. , A transformation to reduce the impact of major peaks compared to minor peaks may be helpful when the compounds of interest are likely to be extremely low in abundance, but not necessarily in the case of natural products discovery.…”

“…Multiple studies have been conducted to evaluate the influence of data processing treatments on subsequent data analysis and have revealed that there are often complex interactions between the parameters used. , To optimize data treatment parameters, it is important to inspect interactions between processing steps. Indeed, statistical analyses also revealed a strong interaction between two data processing steps: data transformation and model simplification using a percent variance cutoff (Figure B and D).…”

Opportunities and Limitations for Untargeted Mass Spectrometry Metabolomics to Identify Biologically Active Constituents in Complex Natural Product Mixtures

“…As the nontransformed data sets were able to identify magnolol as active, the false positives associated with magnolol also increased. These results are counter to the findings of other studies . For example, while Arneberg et al found that the n th root transformation positively impacted their models, models built using this transformation were unable to identify active constituents in the current study.…”

“…These results are counter to the findings of other studies . For example, while Arneberg et al found that the n th root transformation positively impacted their models, models built using this transformation were unable to identify active constituents in the current study. These differences may be due to the differences in applications between these two projects.…”

“…It is common practice in metabolomics studies, particularly those utilizing mass spectrometric data, to subject data to a transformation procedure. , Since mass spectrometers are so sensitive in their ability to detect compounds at a wide range of concentrations, they are subject to errors caused by heteroscedastic noise in count data, in which error is proportional to the peak area. , As such, data transformation processes aimed to reduce the error associated with large peak areas are commonly employed. , Many metabolomics projects utilize, for example, a fourth-root transformation of variable peak areas to minimize the impact of heteroscedastic noise and reduce bias towards highly abundant or ionizable compounds. ,, Despite the popularity of this approach, the statistical analysis revealed that this transformation negatively impacted the ability of models to accurately predict active compounds. Models built using transformed data (Figure E–H) gave berberine and magnolol worse selectivity ratio rankings than data sets using nontransformed data (Figure A–D).…”

“…These differences may be due to the differences in applications between these two projects. While Arneberg et al were assessing proteomics data sets, the data sets assessed for this project were focused on metabolomics-driven natural products discovery. In natural products discovery projects, low-abundant constituents that contribute to bioactivity may be present in the upper parts per million or parts per thousand range, while protein biomarkers are often found in the lower parts per billion range. , A transformation to reduce the impact of major peaks compared to minor peaks may be helpful when the compounds of interest are likely to be extremely low in abundance, but not necessarily in the case of natural products discovery.…”

“…Multiple studies have been conducted to evaluate the influence of data processing treatments on subsequent data analysis and have revealed that there are often complex interactions between the parameters used. , To optimize data treatment parameters, it is important to inspect interactions between processing steps. Indeed, statistical analyses also revealed a strong interaction between two data processing steps: data transformation and model simplification using a percent variance cutoff (Figure B and D).…”

Opportunities and Limitations for Untargeted Mass Spectrometry Metabolomics to Identify Biologically Active Constituents in Complex Natural Product Mixtures

“…As the nontransformed data sets were able to identify magnolol as active, the false positives associated with magnolol also increased. These results are counter to the findings of other studies . For example, while Arneberg et al found that the n th root transformation positively impacted their models, models built using this transformation were unable to identify active constituents in the current study.…”

“…These results are counter to the findings of other studies . For example, while Arneberg et al found that the n th root transformation positively impacted their models, models built using this transformation were unable to identify active constituents in the current study. These differences may be due to the differences in applications between these two projects.…”

“…It is common practice in metabolomics studies, particularly those utilizing mass spectrometric data, to subject data to a transformation procedure. , Since mass spectrometers are so sensitive in their ability to detect compounds at a wide range of concentrations, they are subject to errors caused by heteroscedastic noise in count data, in which error is proportional to the peak area. , As such, data transformation processes aimed to reduce the error associated with large peak areas are commonly employed. , Many metabolomics projects utilize, for example, a fourth-root transformation of variable peak areas to minimize the impact of heteroscedastic noise and reduce bias towards highly abundant or ionizable compounds. ,, Despite the popularity of this approach, the statistical analysis revealed that this transformation negatively impacted the ability of models to accurately predict active compounds. Models built using transformed data (Figure E–H) gave berberine and magnolol worse selectivity ratio rankings than data sets using nontransformed data (Figure A–D).…”

“…These differences may be due to the differences in applications between these two projects. While Arneberg et al were assessing proteomics data sets, the data sets assessed for this project were focused on metabolomics-driven natural products discovery. In natural products discovery projects, low-abundant constituents that contribute to bioactivity may be present in the upper parts per million or parts per thousand range, while protein biomarkers are often found in the lower parts per billion range. , A transformation to reduce the impact of major peaks compared to minor peaks may be helpful when the compounds of interest are likely to be extremely low in abundance, but not necessarily in the case of natural products discovery.…”

“…Multiple studies have been conducted to evaluate the influence of data processing treatments on subsequent data analysis and have revealed that there are often complex interactions between the parameters used. , To optimize data treatment parameters, it is important to inspect interactions between processing steps. Indeed, statistical analyses also revealed a strong interaction between two data processing steps: data transformation and model simplification using a percent variance cutoff (Figure B and D).…”

Opportunities and Limitations for Untargeted Mass Spectrometry Metabolomics to Identify Biologically Active Constituents in Complex Natural Product Mixtures

“…As the nontransformed data sets were able to identify magnolol as active, the false positives associated with magnolol also increased. These results are counter to the findings of other studies . For example, while Arneberg et al found that the n th root transformation positively impacted their models, models built using this transformation were unable to identify active constituents in the current study.…”

“…These results are counter to the findings of other studies . For example, while Arneberg et al found that the n th root transformation positively impacted their models, models built using this transformation were unable to identify active constituents in the current study. These differences may be due to the differences in applications between these two projects.…”

“…It is common practice in metabolomics studies, particularly those utilizing mass spectrometric data, to subject data to a transformation procedure. , Since mass spectrometers are so sensitive in their ability to detect compounds at a wide range of concentrations, they are subject to errors caused by heteroscedastic noise in count data, in which error is proportional to the peak area. , As such, data transformation processes aimed to reduce the error associated with large peak areas are commonly employed. , Many metabolomics projects utilize, for example, a fourth-root transformation of variable peak areas to minimize the impact of heteroscedastic noise and reduce bias towards highly abundant or ionizable compounds. ,, Despite the popularity of this approach, the statistical analysis revealed that this transformation negatively impacted the ability of models to accurately predict active compounds. Models built using transformed data (Figure E–H) gave berberine and magnolol worse selectivity ratio rankings than data sets using nontransformed data (Figure A–D).…”

“…These differences may be due to the differences in applications between these two projects. While Arneberg et al were assessing proteomics data sets, the data sets assessed for this project were focused on metabolomics-driven natural products discovery. In natural products discovery projects, low-abundant constituents that contribute to bioactivity may be present in the upper parts per million or parts per thousand range, while protein biomarkers are often found in the lower parts per billion range. , A transformation to reduce the impact of major peaks compared to minor peaks may be helpful when the compounds of interest are likely to be extremely low in abundance, but not necessarily in the case of natural products discovery.…”

“…Multiple studies have been conducted to evaluate the influence of data processing treatments on subsequent data analysis and have revealed that there are often complex interactions between the parameters used. , To optimize data treatment parameters, it is important to inspect interactions between processing steps. Indeed, statistical analyses also revealed a strong interaction between two data processing steps: data transformation and model simplification using a percent variance cutoff (Figure B and D).…”