Repeat the whole analysis after filtering the full data

[jorainer]

We should repeat the analysis after filtering the data removing potentially poor quality features. Potential criteria could be:

• RSD in QC samples > 30%
• weights/detected calls

[jorainer]

Also suggested by @BeppePaglia .

[jorainer]

The summary for the analysis based on features with an RSD across QC samples < 30% (@BeppePaglia @cvolani @SorenFjel):

Prefiltering

In negative polarity 6,313 of the 16,109 features passed the RSD QC filter, in positive polarity 3,963 of the 10,741. That's 39 and 37% of the features, respectively. All further data analysis bases on these features. Missing values were replaced by half of the feature's smallest intensity adding a small random value.

Matrix specific effects

Qualitative analysis

• Bases on the analysis whether a feature was detected in most samples of a matrix.
• Most features were detected in venous samples both for negative (5,848; 92%) and positive (3,565; 90%) polarity, fewest in capillary (4,392; 69%) in negative polarity and plasma (2,034; 51%) in positive polarity.
• 2,740 (43%) and 1,301 (32%) features are common to/detected in all 4 matrices only between 150 and 50 are matrix specific (i.e. only detected in a single matrix, but not in any other).

Quantitative analysis

• Comparison feature abundances between matrices. Capillary samples are used as baseline, thus plasma, venous and RBC samples are each compared to capillary samples.
• Requirement for differential abundance: at least two-fold difference in abundance and adjusted p-value < 0.01 (i.e. 1% FDR).
• plasma vs capillary: 3,157 (50%) and 2,624 (66%) significant features in negative and positive polarity, respectively.
• RBC vs capillary: 2,376 (37%) and 1,242 (31%) significant features in negative and positive polarity, respectively.
• venous vs capillary: 2,281 (36%) and 1,150 (29%) significant features in negative and positive polarity, respectively.
• venous and capillary samples are thus most similar.
• Distribution of m/z and retention time values is ~ similar for significant features of all 3 comparisons, with the exception of features that have significantly higher abundances in capillary samples than in plasma samples: they seem to have, relative to the other comparison, lower m/z values in negative polarity, but higher m/z in positive polarity. See figures below.

Distribution of m/z (left) and retention time (right) for significant features in negative polarity. Features with lower abundance in plasma than in capillary samples have mostly smaller m/z (between 200 to 400).

Distribution of m/z (left) and retention time (right) for significant features in positive polarity. Features with lower abundance in plasma than in capillary samples have mostly larger m/z (between 600 to 1000).

Sex difference

• Significance criteria was relaxed based on the observation that extent of differences is only moderate: required fold change: 1.62 at a 10% FDR.
• Analysis performed separately for each matrix seems to be better (less influenced by potential confounding effects from other matrices).
• In negative polarity 3 features were found significant in capillary samples and 17 in venous samples with 2 of these features being common to both matrices. In positive polarity only 3 features were significant in plasma samples.
• Comparison of the overlap of the 50 features with the smallest p-value per matrix: 16 features common between capillary and venous samples in negative polarity. Based on their retention times, these could even be adducts of the same 3 compounds. Almost no overlap for top 50 features in plasma or RBC samples. A smaller overlap could be seen for positive polarity: only 8 features common to venous, capillary (and partially also RBC). With the exception of one feature, all top 50 features in plasma are specific for that matrix.

Will provide the reports and xls sheets tomorrow.

[cvolani]

Thanks a lot Johannes, It’s beginning to get a real shape 😊 I’ll wait for your reports and others before making any comments.

Have a nice evening. Best Chiara

Da: Johannes Rainer notifications@github.com Risposta: EuracBiomedicalResearch/VAMS_vs_intraveinous reply@reply.github.com Data: martedì 27 agosto 2019 15:55 A: EuracBiomedicalResearch/VAMS_vs_intraveinous VAMS_vs_intraveinous@noreply.github.com Cc: Volani Chiara Chiara.Volani@eurac.edu, Mention mention@noreply.github.com Oggetto: Re: [EuracBiomedicalResearch/VAMS_vs_intraveinous] Repeat the whole analysis after filtering the full data (#42)

The summary for the analysis based on features with an RSD across QC samples < 30% (@BeppePagliahttps://github.com/BeppePaglia @cvolanihttps://github.com/cvolani @SorenFjelhttps://github.com/SorenFjel):

Prefiltering

In negative polarity 6,313 of the 16,109 features passed the RSD QC filter, in positive polarity 3,963 of the 10,741. That's 39 and 37% of the features, respectively. All further data analysis bases on these features. Missing values were replaced by half of the feature's smallest intensity adding a small random value.

Matrix specific effects Qualitative analysis

• Bases on the analysis whether a feature was detected in most samples of a matrix.
• Most features were detected in venous samples both for negative (5,848; 92%) and positive (3,565; 90%) polarity, fewest in capillary (4,392; 69%) in negative polarity and plasma (2,034; 51%) in positive polarity.
• 2,740 (43%) and 1,301 (32%) features are common to/detected in all 4 matrices only between 150 and 50 are matrix specific (i.e. only detected in a single matrix, but not in any other).

Quantitative analysis

• Comparison feature abundances between matrices. Capillary samples are used as baseline, thus plasma, venous and RBC samples are each compared to capillary samples.
• Requirement for differential abundance: at least two-fold difference in abundance and adjusted p-value < 0.01 (i.e. 1% FDR).
• plasma vs capillary: 3,157 (50%) and 2,624 (66%) significant features in negative and positive polarity, respectively.
• RBC vs capillary: 2,376 (37%) and 1,242 (31%) significant features in negative and positive polarity, respectively.
• venous vs capillary: 2,281 (36%) and 1,150 (29%) significant features in negative and positive polarity, respectively.
• venous and capillary samples are thus most similar.
• Distribution of m/z and retention time values is ~ similar for significant features of all 3 comparisons, with the exception of features that have significantly higher abundances in capillary samples than in plasma samples: they seem to have, relative to the other comparison, lower m/z values in negative polarity, but higher m/z in positive polarity. See figures below.

[mzCountsSignificant-1]https://user-images.githubusercontent.com/5506112/63776237-44dc5d00-c8e1-11e9-83a8-521b5f3b9e78.png Distribution of m/z (left) and retention time (right) for significant features in negative polarity. Features with lower abundance in plasma than in capillary samples have mostly smaller m/z (between 200 to 400).

[mzCountsSignificant-1]https://user-images.githubusercontent.com/5506112/63776395-9d135f00-c8e1-11e9-84ba-d7e95e2588be.png Distribution of m/z (left) and retention time (right) for significant features in positive polarity. Features with lower abundance in plasma than in capillary samples have mostly larger m/z (between 600 to 1000).

Sex difference

• Significance criteria was relaxed based on the observation that extent of differences is only moderate: required fold change: 1.62 at a 10% FDR.
• Analysis performed separately for each matrix seems to be better (less influenced by potential confounding effects from other matrices).
• In negative polarity 3 features were found significant in capillary samples and 17 in venous samples with 2 of these features being common to both matrices. In positive polarity only 3 features were significant in plasma samples.
• Comparison of the overlap of the 50 features with the smallest p-value per matrix: 16 features common between capillary and venous samples in negative polarity. Based on their retention times, these could even be adducts of the same 3 compounds. Almost no overlap for top 50 features in plasma or RBC samples. A smaller overlap could be seen for positive polarity: only 8 features common to venous, capillary (and partially also RBC). With the exception of one feature, all top 50 features in plasma are specific for that matrix.

Will provide the reports and xls sheets tomorrow.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/EuracBiomedicalResearch/VAMS_vs_intraveinous/issues/42?email_source=notifications&email_token=AKXB2ZSSDQJYEEYFQK4VZGDQGUW4ZA5CNFSM4HPD4O6KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5H2FFA#issuecomment-525312660, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AKXB2ZSQUQQXGRNNN2C2MZDQGUW4ZANCNFSM4HPD4O6A.