Metabonomics SOP - Revision history

Rasajdak: /* Calculation of Figures-Of-Merit (FOM) for processed 1D NMR spectra */

2012-03-02T16:14:30Z

Calculation of Figures-Of-Merit (FOM) for processed 1D NMR spectra

@@ Line 393: / Line 393: @@
 === Calculation of Figures-Of-Merit (FOM) for processed 1D NMR spectra  ===
-To ensure (i)&nbsp;unbiased baseline correction, (ii) a high quality of shims and (iii) an accurate phase correction,&nbsp;FOMs were calculated for each spectrum. For (i), the average and standard deviation of data&nbsp;points on the spectral regions containing only baseline (e.g., 14.7–9.0 ppm and –1.0 – -5.2 ppm)&nbsp;were calculated. The cut-offs chosen for the EOC project were: average &lt; 2 x 10–6 and standard&nbsp;deviation &lt; 5 x 10–6. For (ii) and (iii), a Lorentzian line was fitted into the resonance line of the&nbsp;internal standard formate. The residual (that is, the difference between the actual peak shape and&nbsp;the Lorentzian line) was integrated after the fit. The cut-offs chosen were &lt;2% of the total&nbsp;integral of the formate resonance line. Additionally, the symmetry of the formate line was&nbsp;assessed by calculating the difference between peak segments located down- and upfield of the&nbsp;peak maximum. The cut-off chosen for the difference was &lt;4% of the total integral.&lt;font size="4"&lt;/font&gt;
+To ensure (i)&nbsp;unbiased baseline correction, (ii) a high quality of shims and (iii) an accurate phase correction,&nbsp;FOMs were calculated for each spectrum. For (i), the average and standard deviation of data&nbsp;points on the spectral regions containing only baseline (e.g., 14.7–9.0 ppm and –1.0 – -5.2 ppm)&nbsp;were calculated. The cut-offs chosen for the EOC project were: average &lt; 2 x 10–6 and standard&nbsp;deviation &lt; 5 x 10–6. For (ii) and (iii), a Lorentzian line was fitted into the resonance line of the&nbsp;internal standard formate. The residual (that is, the difference between the actual peak shape and&nbsp;the Lorentzian line) was integrated after the fit. The cut-offs chosen were &lt;2% of the total&nbsp;integral of the formate resonance line. Additionally, the symmetry of the formate line was&nbsp;assessed by calculating the difference between peak segments located down- and upfield of the&nbsp;peak maximum. The cut-off chosen for the difference was &lt;4% of the total integral.
 == Operator Certification  ==

Rasajdak at 14:53, 1 March 2012

2012-03-01T14:53:56Z

← Older revision		Revision as of 14:53, 1 March 2012
Line 429:		Line 429:
	== Development of Metabolic Models ==		== Development of Metabolic Models ==

	== Data Archiving ==		== Data Archiving ==


			Since metabonomics involves handling of a large number of raw and processed data, it is essential to maintain a good ‘bookkeeping’ and data back-ups.

			#Take the sample list (i.e. excel file) from the Sample Storage Protocol and record the raw FID name for each NMR experiment along with the run order numbers.
			#Record the filename of the corresponding processed spectra beside the raw FID column.
			#When generating processed spectra for multivariate data analysis (in the form of N x K matrix), indicate the project name and whether the spectra are binned or unbinned in the filename. Record the data description in your lab journal. Additionally, generate a list of sample filename and info/description for each specimen and upload it to the HTP server (/nsm/chem/cen2/HTP2/1_projects/METABONOMICS).
			##After data collection, the raw FID should be saved in the spectrometer computer and lab desktop as soon as possible.
			##The raw FID and the processed spectra should be saved in at least one of the lab desktops (e.g. meta (NSC 820), moscow (NSC 820) and spins3 (NSC 807)), HTP server and a separate disk.

	= MS =		= MS =

	= Statistics =		= Statistics =

Rasajdak at 14:52, 1 March 2012

2012-03-01T14:52:47Z

← Older revision		Revision as of 14:52, 1 March 2012
Line 427:		Line 427:
	#Spike the compounds as ranked in 9 to obtain ultimate confirmation of resonance assignments.		#Spike the compounds as ranked in 9 to obtain ultimate confirmation of resonance assignments.

	== Development of Metabolic Models~~[http://nmr2.buffalo.edu/bufweb/tiki-editpage.php?page~~=~~Development+of+Metabolic+Models ?]~~ ==		== Development of Metabolic Models ==

			== Data Archiving ==

	= MS =		= MS =

	= Statistics =		= Statistics =

Rasajdak: /* Identification of metabolites and associated pathways that are affected upon onset of a disease */

2012-03-01T14:51:34Z

Identification of metabolites and associated pathways that are affected upon onset of a disease

← Older revision		Revision as of 14:51, 1 March 2012
Line 377:		Line 377:
	#Record and report all potential outliers and justification for exclusion, modification or neither. The summary should indicate which spectra are deemed as outliers due to the presence of resonance lines of unknown identity and for reasons stated in (1 above).		#Record and report all potential outliers and justification for exclusion, modification or neither. The summary should indicate which spectra are deemed as outliers due to the presence of resonance lines of unknown identity and for reasons stated in (1 above).

	== Identification of metabolites and associated pathways that are affected upon onset of a disease ==		== Identification of metabolites and associated pathways that are affected upon onset of a disease ==

	After NMR data collection, processing and statistical analysis, the information from NMR and statistics are interpreted into the context of biological changes in the human body upon onset of a disease. The step-wise procedure towards building a metabolic hypothesis from changes in the NMR profile of metabolites present in serum is outlined below.		After NMR data collection, processing and statistical analysis, the information from NMR and statistics are interpreted into the context of biological changes in the human body upon onset of a disease. The step-wise procedure towards building a metabolic hypothesis from changes in the NMR profile of metabolites present in serum is outlined below.
	#Determine the spectral regions of 1D 1H DIRE, 1D 1H DOSY, skyline projection of 2D Jresolved, 1D 1H CPMG, 1D 1H NOESY, 2D J-resolved and 2D [1H,1H]-TOCSY spectra
	that are significantly different (p-value < 0.05) between spectra recorded for specimens from ‘healthy controls’ and the ‘diseased’ group.		#Determine the spectral regions of 1D 1H DIRE, 1D 1H DOSY, skyline projection of 2D Jresolved, 1D 1H CPMG, 1D 1H NOESY, 2D J-resolved and 2D [1H,1H]-TOCSY spectra that are significantly different (p-value < 0.05) between spectra recorded for specimens from ‘healthy controls’ and the ‘diseased’ group.
	#Perform resonance assignment for the spectral regions identified in step (1). Metabolites with p-values < 0.05 will be referred to as ‘significant’ metabolites from here onwards.		#Perform resonance assignment for the spectral regions identified in step (1). Metabolites with p-values < 0.05 will be referred to as ‘significant’ metabolites from here onwards.
	#Determine which metabolites in step (2) increase or decrease in spectra of ‘diseased’ relative to ‘healthy controls’.		#Determine which metabolites in step (2) increase or decrease in spectra of ‘diseased’ relative to ‘healthy controls’.
	#Color-code and rank the ‘significant’ metabolites according to p-values. Use the following colors to indicate degree of significance: black ≤ ~~10–5~~, violet = ~~10–4~~, green = ~~10–3~~ and orange = 0.05 – 0.01.		#Color-code and rank the ‘significant’ metabolites according to p-values. Use the following colors to indicate degree of significance: black ≤ 10<sup>–5</sup>, violet = 10<sup>–4</sup>, green = 10<sup>–3</sup> and orange = 0.05 – 0.01.
	#Map the ‘significant’ metabolites onto human metabolic pathways which can		#Map the ‘significant’ metabolites onto human metabolic pathways which can be obtained online from the Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.jp), starting with those that are colored black and violet prior to those that are colored green and orange.
	be obtained online from the Kyoto Encyclopedia of Genes and Genomes (KEGG;		#Complement the resulting metabolic chart from step (5) with information on the location of pathways perturbed with the onset of the disease and the observations regarding tumor metabolism from literature. The site of each metabolic pathway (e.g., organelle, cell, tissue, organ) can be obtained from the Human Metabolome Database (HMDB; http://www.hmdb.ca/).
	http://www.genome.jp), starting with those that are colored black and violet prior to those		#Compare the changes in NMR profile of metabolites in serum with the metabolite profile for the disease obtained from tissue/tumor biopsies. This could be used to determine if the changes in NMR profiles observed in serum is a systemic response of the human body to the damaged organ/tumor.
	that are colored green and orange.
	#Complement the resulting metabolic chart from step (5) with information on the location
	of pathways perturbed with the onset of the disease and the observations regarding tumor
	metabolism from literature. The site of each metabolic pathway (e.g., organelle, cell,
	tissue, organ) can be obtained from the Human Metabolome Database (HMDB;
	http://www.hmdb.ca/).
	#Compare the changes in NMR profile of metabolites in serum with the metabolite profile
	for the disease obtained from tissue/tumor biopsies. This could be used to determine if
	the changes in NMR profiles observed in serum is a systemic response of the human
	body to the damaged organ/tumor.

	== Spectral Quality Assessment (FOM) ==		== Spectral Quality Assessment (FOM) ==

Rasajdak at 14:50, 1 March 2012

2012-03-01T14:50:04Z

@@ Line 376: / Line 376: @@
 #Repeat steps (4) – (6) for the rest of the observations on the table from (2). Using the modified excel/csv file, redo the MVDA. Repeat outlier identification process after performing MVDA.
 #Record and report all potential outliers and justification for exclusion, modification or neither. The summary should indicate which spectra are deemed as outliers due to the presence of resonance lines of unknown identity and for reasons stated in (1 above).
 == Spectral Quality Assessment (FOM)  ==

Rasajdak: /* Outlier Identification */

2012-03-01T01:31:10Z

Outlier Identification

← Older revision		Revision as of 01:31, 1 March 2012
Line 373:		Line 373:
	#Repeat steps (1)-(2) on PCs 4 to 10 (This number of PCs considered is for a large number of samples N) pairwise.		#Repeat steps (1)-(2) on PCs 4 to 10 (This number of PCs considered is for a large number of samples N) pairwise.
	#Check the score contribution plot of each observation on the table generated according to (2). The contribution plot (Figure 3.2.8.A) displays the differences, in scaled units, for all the variables in the model, between a specific spectrum and the average spectrum, multiplied by the absolute value of the normalized weight (SIMCA-P+ Manual.Version 11.0. 2005. UMETRICS, Sweden). In other words, the contribution plot of an observation explains why it is different from most of the samples. List the bins with the most significant contribution (most intense).		#Check the score contribution plot of each observation on the table generated according to (2). The contribution plot (Figure 3.2.8.A) displays the differences, in scaled units, for all the variables in the model, between a specific spectrum and the average spectrum, multiplied by the absolute value of the normalized weight (SIMCA-P+ Manual.Version 11.0. 2005. UMETRICS, Sweden). In other words, the contribution plot of an observation explains why it is different from most of the samples. List the bins with the most significant contribution (most intense).
	#For each sample, check the peak(s) corresponding to the bin(s) identified in (7) on the processed 1D NMR spectrum. If the peak(s) is(are) isolated and judged as anomalous (refer to Figure 3.2.8.B), replace the bin intensities in the excel/csv file with the median value (of other spectra). Otherwise, exclude the spectrum (Figure 3.2.8.C shows an example of an ‘outlier’ excluded for MVDA).		#For each sample, check the peak(s) corresponding to the bin(s) identified in (4) on the processed 1D NMR spectrum. If the peak(s) is(are) isolated and judged as anomalous (refer to Figure 3.2.8.B), replace the bin intensities in the excel/csv file with the median value (of other spectra). Otherwise, exclude the spectrum (Figure 3.2.8.C shows an example of an ‘outlier’ excluded for MVDA).
			#Repeat steps (4) – (6) for the rest of the observations on the table from (2). Using the modified excel/csv file, redo the MVDA. Repeat outlier identification process after performing MVDA.
			#Record and report all potential outliers and justification for exclusion, modification or neither. The summary should indicate which spectra are deemed as outliers due to the presence of resonance lines of unknown identity and for reasons stated in (1 above).

	== Spectral Quality Assessment (FOM) ==		== Spectral Quality Assessment (FOM) ==

Rasajdak: /* Outlier Identification */

2012-03-01T01:29:47Z

Outlier Identification

← Older revision		Revision as of 01:29, 1 March 2012
Line 372:		Line 372:
	#List the observations which are classified as outliers based on both 95%-ellipse and DModX.		#List the observations which are classified as outliers based on both 95%-ellipse and DModX.
	#Repeat steps (1)-(2) on PCs 4 to 10 (This number of PCs considered is for a large number of samples N) pairwise.		#Repeat steps (1)-(2) on PCs 4 to 10 (This number of PCs considered is for a large number of samples N) pairwise.
	#Check the score contribution plot of each observation on the table generated according to (2). The contribution plot (Figure 3.2.8.A) displays the differences, in scaled units, for all the variables in the model, between a specific spectrum and the average spectrum, multiplied by the absolute value of the normalized weight (SIMCA-P+ Manual.Version 11.0. 2005. UMETRICS, Sweden).		#Check the score contribution plot of each observation on the table generated according to (2). The contribution plot (Figure 3.2.8.A) displays the differences, in scaled units, for all the variables in the model, between a specific spectrum and the average spectrum, multiplied by the absolute value of the normalized weight (SIMCA-P+ Manual.Version 11.0. 2005. UMETRICS, Sweden). In other words, the contribution plot of an observation explains why it is different from most of the samples. List the bins with the most significant contribution (most intense).
			#For each sample, check the peak(s) corresponding to the bin(s) identified in (7) on the processed 1D NMR spectrum. If the peak(s) is(are) isolated and judged as anomalous (refer to Figure 3.2.8.B), replace the bin intensities in the excel/csv file with the median value (of other spectra). Otherwise, exclude the spectrum (Figure 3.2.8.C shows an example of an ‘outlier’ excluded for MVDA).

	== Spectral Quality Assessment (FOM) ==		== Spectral Quality Assessment (FOM) ==

Rasajdak: /* Outlier Identification */

2012-03-01T01:29:03Z

Outlier Identification

← Older revision		Revision as of 01:29, 1 March 2012
Line 367:		Line 367:
	#Identify spectra with anomalous (e.g. very intense and/or broad) peaks and spectra with peaks on unexpected positions. Check spectra individually and judge if each is to be excluded.		#Identify spectra with anomalous (e.g. very intense and/or broad) peaks and spectra with peaks on unexpected positions. Check spectra individually and judge if each is to be excluded.

	""After MVDA""		'''After MVDA'''

	#Plot PC1 vs PC2 (Pareto-scaled, bin size = 0.005ppm) and generate a Distance-to-Model (DModX) plot. Tabulate the observations (samples) which are outside the 95%-ellipse. Rank them according to distance from the ellipse. Repeat this for DModX.		#Plot PC1 vs PC2 (Pareto-scaled, bin size = 0.005ppm) and generate a Distance-to-Model (DModX) plot. Tabulate the observations (samples) which are outside the 95%-ellipse. Rank them according to distance from the ellipse. Repeat this for DModX.

Rasajdak: /* Outlier Identification */

2012-03-01T01:28:33Z

Outlier Identification

← Older revision		Revision as of 01:28, 1 March 2012
Line 368:		Line 368:

	""After MVDA""		""After MVDA""

	#Plot PC1 vs PC2 (Pareto-scaled, bin size = 0.005ppm) and generate a Distance-to-Model (DModX) plot. Tabulate the observations (samples) which are outside the 95%-ellipse. Rank them according to distance from the ellipse. Repeat this for DModX.		#Plot PC1 vs PC2 (Pareto-scaled, bin size = 0.005ppm) and generate a Distance-to-Model (DModX) plot. Tabulate the observations (samples) which are outside the 95%-ellipse. Rank them according to distance from the ellipse. Repeat this for DModX.
	#List the observations which are classified as outliers based on both 95%-ellipse and DModX.		#List the observations which are classified as outliers based on both 95%-ellipse and DModX.
	#Repeat steps (4)-(5) on PCs 4 to 10 (This number of PCs considered is for a large number of samples N) pairwise.		#Repeat steps (1)-(2) on PCs 4 to 10 (This number of PCs considered is for a large number of samples N) pairwise.
	#Check the score contribution plot of each observation on the table generated according to (5). The contribution plot (Figure 3.2.8.A) displays the differences, in scaled units, for all the variables in the model, between a specific spectrum and the average spectrum, multiplied by the absolute value of the normalized weight (SIMCA-P+ Manual.Version 11.0. 2005. UMETRICS, Sweden).		#Check the score contribution plot of each observation on the table generated according to (2). The contribution plot (Figure 3.2.8.A) displays the differences, in scaled units, for all the variables in the model, between a specific spectrum and the average spectrum, multiplied by the absolute value of the normalized weight (SIMCA-P+ Manual.Version 11.0. 2005. UMETRICS, Sweden).

	== Spectral Quality Assessment (FOM) ==		== Spectral Quality Assessment (FOM) ==

Rasajdak: /* Data Processing */

2012-03-01T01:27:52Z

Data Processing

← Older revision		Revision as of 01:27, 1 March 2012
Line 356:		Line 356:

	The time domain data were zero-filled to 16384 and 2048 points along F2 and F1 dimensions, respectively. A cosine-bell squared window function was applied to F2 and a cosine-bell window function was applied to F1 dimension prior to FT. Each 2D spectrum was calibrated to the formate resonance line at (8.444, 172.0) ppm after phase and baseline correction.		The time domain data were zero-filled to 16384 and 2048 points along F2 and F1 dimensions, respectively. A cosine-bell squared window function was applied to F2 and a cosine-bell window function was applied to F1 dimension prior to FT. Each 2D spectrum was calibrated to the formate resonance line at (8.444, 172.0) ppm after phase and baseline correction.

			== Outlier Identification ==

			The initial criteria for selection of outliers is based on S/N and line width for the internal standard formate, quality of water suppression and spectral processing artifacts. This is followed by a more thorough outlier identification steps by Principal Component Analysis (PCA) scores plot and Distance-to-Model plot. The details are shown below.

			'''Before Multi Variate Data Analysis (MVDA)'''

			#After each batch of data acquired, perform a quality assessment of each 1D NMR spectrum (as indicated in SOP for serum sample data acquisition). Exclude the spectra which have unacceptable formate H1 resonance S/N and linewidth. Additionally, exclude spectra with unacceptable water suppression.
			#Overlay all 1D-NOESY (or 1D CPMG) spectra and expand the formate region to check if the calibration is accurate. Scroll over the overlay to check for obvious phase and baseline errors, and by rephrasing, recalibrating, and/or repeating the baseline correction.
			#Identify spectra with anomalous (e.g. very intense and/or broad) peaks and spectra with peaks on unexpected positions. Check spectra individually and judge if each is to be excluded.

			""After MVDA""
			#Plot PC1 vs PC2 (Pareto-scaled, bin size = 0.005ppm) and generate a Distance-to-Model (DModX) plot. Tabulate the observations (samples) which are outside the 95%-ellipse. Rank them according to distance from the ellipse. Repeat this for DModX.
			#List the observations which are classified as outliers based on both 95%-ellipse and DModX.
			#Repeat steps (4)-(5) on PCs 4 to 10 (This number of PCs considered is for a large number of samples N) pairwise.
			#Check the score contribution plot of each observation on the table generated according to (5). The contribution plot (Figure 3.2.8.A) displays the differences, in scaled units, for all the variables in the model, between a specific spectrum and the average spectrum, multiplied by the absolute value of the normalized weight (SIMCA-P+ Manual.Version 11.0. 2005. UMETRICS, Sweden).

	== Spectral Quality Assessment (FOM) ==		== Spectral Quality Assessment (FOM) ==