1. Metabolomic analysis of plasma samples from obese African American women identified lipid markers that can predict type 2 diabetes (T2D) status and polymorphisms in the UCP3 gene.
2. Certain monoacylglycerols, N-acylethanolamides, and 18-carbon epoxides were increased in T2D and correlated with fasting glucose levels, while 5-lipoxygenase activity was decreased.
3. The effects of T2D on lipid markers were modified by UCP3 genotype, with linoleate di- and tri-hydroxyls and steroylethanolamide increased in the UCP3 g/a genotype compared to the g/
general information regarding single nucleotide polymorphism.
A Single Nucleotide Polymorphisms (SNP), pronounced “snip,” is a genetic variation when a single nucleotide (i.e., A, T, C, or G) is altered and kept through heredity.
general information regarding single nucleotide polymorphism.
A Single Nucleotide Polymorphisms (SNP), pronounced “snip,” is a genetic variation when a single nucleotide (i.e., A, T, C, or G) is altered and kept through heredity.
Poster - RNAi Therapeutics by Silence Therapeutics - American Society of Cli...Silence Therapeutics
A poster detailing the results from the first-in-human Phase I study of Atu027, novel RNAi therapeutic. Atu027 is a liposomal siRNA formulation that targets protein kinase N3 (PKN3) in patients with advanced solid tumors. Find out more here:
http://silence-therapeutics.com/content/pipeline/internalprograms.htm
Cell and gene therapy for Parkinson’s disease - part 2Parkinson's UK
Presentation by Prof Deniz Kirik, MD, PhD at the Parkinson's UK Research Conference, November 2010 in York.
With introduction by Dr Oliver Bandmann.
Part 1: http://www.slideshare.net/ParkinsonsResearchUK/cell-and-gene-therapy-for-parkinsons-disease-part-1
Full course: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/
The course covered all of the steps required to go from `raw data` to a rich `mapped biochemical network` incorporating statistical, multivariate and machine learning results. This included [examples](https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/#topics) and tutorials for:
* Preparing raw data for analysis
* Multivariate data exploration
* Supervised clustering
* Machine learning – classification model validation and feature selection
* Network analysis - biochemical, structural similarity and correlation networks
* Network mapping – putting it all together to create a publication quality network
url:
https://github.com/CreativeDataSolutions/CDS.courses/blob/gh-pages/courses/network_mapping_101/materials/lectures/tutorial.pdf
Rise of Deep Learning for Genomic, Proteomic, and Metabolomic Data Integratio...Dmitry Grapov
Machine learning (ML) is being ubiquitously incorporated into everyday products such as Internet search, email spam filters, product recommendations, image classification, and speech recognition. New approaches for highly integrated manufacturing and automation such as the Industry 4.0 and the Internet of things are also converging with ML methodologies. Many approaches incorporate complex artificial neural network architectures and are collectively referred to as deep learning (DL) applications. These methods have been shown capable of representing and learning predictable relationships in many diverse forms of data and hold promise for transforming the future of omics research and applications in precision medicine. Omics and electronic health record data pose considerable challenges for DL. This is due to many factors such as low signal to noise, analytical variance, and complex data integration requirements. However, DL models have already been shown capable of both improving the ease of data encoding and predictive model performance over alternative approaches. It may not be surprising that concepts encountered in DL share similarities with those observed in biological message relay systems such as gene, protein, and metabolite networks. This expert review examines the challenges and opportunities for DL at a systems and biological scale for a precision medicine readership.
current: https://drive.google.com/open?id=0B51AEMfo-fh9M3FmWXVlb05pdm8
I am always looking for the next data science, machine learning and visualization challenge.
Here is a link to my up to date
resume:
https://drive.google.com/open?id=0B51AEMfo-fh9M3FmWXVlb05pdm8
cv:
https://drive.google.com/open?id=0B51AEMfo-fh9Z05aM2p6XzFIOFE
https://www.youtube.com/watch?v=Y_-o-4rKxUk
Machine learning powered metabolomic network analysis
Dmitry Grapov PhD,
Director of Data Science and Bioinformatics,
CDS- Creative Data Solutions
www.createdatasol.com
Metabolomic network analysis can be used to interpret experimental results within a variety of contexts including: biochemical relationships, structural and spectral similarity and empirical correlation. Machine learning is useful for modeling relationships in the context of pattern recognition, clustering, classification and regression based predictive modeling. The combination of developed metabolomic networks and machine learning based predictive models offer a unique method to visualize empirical relationships while testing key experimental hypotheses. The following presentation focuses on data analysis, visualization, machine learning and network mapping approaches used to create richly mapped metabolomic networks. Learn more at www.createdatasol.com
Case Study: Overview of Metabolomic Data Normalization StrategiesDmitry Grapov
Five normalization methods were compared, of which the combination of qc-LOESS and cubic splines showed the best performance based on within-batch and between-batch variable relative standard deviations for QCs. This approach was used to normalize sample measurements the results of which were analyzed using principal components analysis.
Poster - RNAi Therapeutics by Silence Therapeutics - American Society of Cli...Silence Therapeutics
A poster detailing the results from the first-in-human Phase I study of Atu027, novel RNAi therapeutic. Atu027 is a liposomal siRNA formulation that targets protein kinase N3 (PKN3) in patients with advanced solid tumors. Find out more here:
http://silence-therapeutics.com/content/pipeline/internalprograms.htm
Cell and gene therapy for Parkinson’s disease - part 2Parkinson's UK
Presentation by Prof Deniz Kirik, MD, PhD at the Parkinson's UK Research Conference, November 2010 in York.
With introduction by Dr Oliver Bandmann.
Part 1: http://www.slideshare.net/ParkinsonsResearchUK/cell-and-gene-therapy-for-parkinsons-disease-part-1
Similar to Diabetes and Obesity Conference, Keystone 2011 (11)
Full course: https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/
The course covered all of the steps required to go from `raw data` to a rich `mapped biochemical network` incorporating statistical, multivariate and machine learning results. This included [examples](https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/#topics) and tutorials for:
* Preparing raw data for analysis
* Multivariate data exploration
* Supervised clustering
* Machine learning – classification model validation and feature selection
* Network analysis - biochemical, structural similarity and correlation networks
* Network mapping – putting it all together to create a publication quality network
url:
https://github.com/CreativeDataSolutions/CDS.courses/blob/gh-pages/courses/network_mapping_101/materials/lectures/tutorial.pdf
Rise of Deep Learning for Genomic, Proteomic, and Metabolomic Data Integratio...Dmitry Grapov
Machine learning (ML) is being ubiquitously incorporated into everyday products such as Internet search, email spam filters, product recommendations, image classification, and speech recognition. New approaches for highly integrated manufacturing and automation such as the Industry 4.0 and the Internet of things are also converging with ML methodologies. Many approaches incorporate complex artificial neural network architectures and are collectively referred to as deep learning (DL) applications. These methods have been shown capable of representing and learning predictable relationships in many diverse forms of data and hold promise for transforming the future of omics research and applications in precision medicine. Omics and electronic health record data pose considerable challenges for DL. This is due to many factors such as low signal to noise, analytical variance, and complex data integration requirements. However, DL models have already been shown capable of both improving the ease of data encoding and predictive model performance over alternative approaches. It may not be surprising that concepts encountered in DL share similarities with those observed in biological message relay systems such as gene, protein, and metabolite networks. This expert review examines the challenges and opportunities for DL at a systems and biological scale for a precision medicine readership.
current: https://drive.google.com/open?id=0B51AEMfo-fh9M3FmWXVlb05pdm8
I am always looking for the next data science, machine learning and visualization challenge.
Here is a link to my up to date
resume:
https://drive.google.com/open?id=0B51AEMfo-fh9M3FmWXVlb05pdm8
cv:
https://drive.google.com/open?id=0B51AEMfo-fh9Z05aM2p6XzFIOFE
https://www.youtube.com/watch?v=Y_-o-4rKxUk
Machine learning powered metabolomic network analysis
Dmitry Grapov PhD,
Director of Data Science and Bioinformatics,
CDS- Creative Data Solutions
www.createdatasol.com
Metabolomic network analysis can be used to interpret experimental results within a variety of contexts including: biochemical relationships, structural and spectral similarity and empirical correlation. Machine learning is useful for modeling relationships in the context of pattern recognition, clustering, classification and regression based predictive modeling. The combination of developed metabolomic networks and machine learning based predictive models offer a unique method to visualize empirical relationships while testing key experimental hypotheses. The following presentation focuses on data analysis, visualization, machine learning and network mapping approaches used to create richly mapped metabolomic networks. Learn more at www.createdatasol.com
Case Study: Overview of Metabolomic Data Normalization StrategiesDmitry Grapov
Five normalization methods were compared, of which the combination of qc-LOESS and cubic splines showed the best performance based on within-batch and between-batch variable relative standard deviations for QCs. This approach was used to normalize sample measurements the results of which were analyzed using principal components analysis.
3 data normalization (2014 lab tutorial)Dmitry Grapov
Get more information:
http://imdevsoftware.wordpress.com/2014/10/11/2014-metabolomic-data-analysis-and-visualization-workshop-and-tutorials/
Recently I had the pleasure of teaching statistical and multivariate data analysis and visualization at the annual Summer Sessions in Metabolomics 2014, organized by the NIH West Coast Metabolomics Center.
Metabolomic Data Analysis Workshop and Tutorials (2014)Dmitry Grapov
Get more information:
http://imdevsoftware.wordpress.com/2014/10/11/2014-metabolomic-data-analysis-and-visualization-workshop-and-tutorials/
Recently I had the pleasure of teaching statistical and multivariate data analysis and visualization at the annual Summer Sessions in Metabolomics 2014, organized by the NIH West Coast Metabolomics Center.
Similar to last year, I’ve posted all the content (lectures, labs and software) for any one to follow along with at their own pace. I also plan to release videos for all the lectures and labs.
Step by step tutorial for conducting GO enrichment analysis and then creating a network from the results.
Material from the UC Davis 2014 Proteomics Workshop.
See more at: http://sourceforge.net/projects/teachingdemos/files/2014%20UC%20Davis%20Proteomics%20Workshop/
Prote-OMIC Data Analysis and VisualizationDmitry Grapov
Introductory lecture to multivariate analysis of proteomic data.
Material from the UC Davis 2014 Proteomics Workshop.
See more at: http://sourceforge.net/projects/teachingdemos/files/2014%20UC%20Davis%20Proteomics%20Workshop/
Data Normalization Approaches for Large-scale Biological StudiesDmitry Grapov
Overview of how to estimate data quality and validate normalization approaches to remove analytical variance.
See here for animations used in the presentation:
http://imdevsoftware.wordpress.com/2014/06/04/using-repeated-measures-to-remove-artifacts-from-longitudinal-data/
Data Normalization Approaches for Large-scale Biological Studies
Diabetes and Obesity Conference, Keystone 2011
1. Bioactive lipid markers of Type 2 Diabetes and UCP3 gene polymorphism in obese
African-American women
Dmitry Grapov1, Sean H. Adams2,3, W. Timothy Garvey4, Kerry H. Lok4,John W. Newman2,3
1Agricultural and Environmental Chemistry, University of California Davis, 2Obesity & Metabolism Research Unit, USDA-ARS Western Human Nutrition
Research Center, Davis, CA, 3Nutrition, University of California Davis, Davis, CA, 4Nutrition Sciences, University of Alabama, Birmingham, AL
Abstract Results Conclusions
Non-esterified free fatty acids (NEFA), endocannabinoids
Only 10% of T2D-dependent metabolic OPLS-DA Models of Entire Population Regression analyses of the OPLS-DA Increased
(eCBs) and oxylipins (OxL) are hypothesized to influence in T2D Type 2 Diabetes Effects
changes are conserved between UCP3 Discriminate T2D but not UCP3 Latent Variable Discriminating T2D
and/or reflect insulin resistance (IR) and lipid/carbohydrate
Genotypes sub-populations
fuel selection. We further hypothesize that the quantitative 2.45
1
g/a UCP3
metabolomic investigation of these lipid classes will provide 2.35
0.8
g/g UCP3
log(Fasting Glucose)
metabolic signatures of organismal fat oxidation and insulin 2.5 2.25
0.6
sensitivity. To test this hypothesis we conducted a study to
2.15
2
All 2
R = 0.508
p<1e-7 2.05
evaluate the ability of these metabolites to report clinical
Loadings 1
1.5 0.4
1.95
diabetes and changes in mitochondrial function in a cohort of 1 1.85
BMI- and age-matched obese African-American women 33 0.5 1.75
0.2
-5 -4 -3 -2 -1 0 1 2 3
containing both diabetic (n=43) and non-diabetic (n=12) 0
23
0 t[1]
6
t[2]
subjects equally distributed between the wildtype (g/g) and -6 -5 -4 -3 -2 -1 0 1 2 3
4 -0.5 Fig 3A: OPLS-DA Latent Variable and the -0.2
missense (g/a) UCP3 genotypes. This study aims to answer 28 7 -1 log of the fasting glucose are linearly MAGs NAEs C18- MUFA/Sat. 5-LOX
epoxides
the following: 4 g/g T2D -1.5 correlated.
-0.4
g/a T2D
1. Can bioactive lipids be used to predict the T2D -2
Summarized effects of T2D:
52
g/g non-T2D
phenotype and UCP3 gene polymorphism? g/g UCP3 g/a UCP3 -2.5 47
g/a non-T2D
2. Do circulating lipid markers of UCP3 polymorphism Significance determined from two-tailed Student’s t-tests (p<0.05 , q=0.1) Test Set -3 42
1. Metabolites correlated to BMI change e.g.:
provide insight into UCP3 function? evaluated on log transformed metabolite concentrations. t[1] 37
monoacylglycerols
BM I
2
R = 0.45
3. Does UCP3 impact the metabolic signature of T2D? p<0.02 32
Fig 1: A Venn Diagram of plasma Fig 2: Scores plot of OPLS-DA models built 27 N-acylethanolamides
Using an array of megavariate statistics we identified metabolites significantly altered in T2D with from a subset of metabolites selected for 22
previously unknown relationships between NEFA, eCBs and UCP3 polymorph interactions. A 62% optimal sub-population discrimination. 17 2. Metabolites marking glucose control increase
18 carbon epoxides (C18-epoxides). These relationships increase in 4 MUFA species occurred in all Physical and clinical parameter driven models
-5 -4 -3 -2 -1
t[1]
0 1 2 3
18 carbon epoxides
only emerge in the T2D state. Changes in these species are T2Ds regardless of UCP3 genotype. (i.e. fasting glucose, BMI, triglycerides) perform Fig 3B: OPLS-DA Latent Variable and MUFA and Sats
shown to be dependent on UCP3 function, suggesting worse than metabolite driven models: 11% v. 7% BMI of non-T2D but not to T2D are
evidence for UCP3’s role in modifying the observed T2D misclassification). linearly correlated. 3. 5-lipoxygenase activity, potentially linked to
phenotype. insulin secretion.
OPLS-DA Models of T2D Effects Among UCP3 Sub-populations UCP3 effects are more evident in non-T2D compared to T2D
6 3
g/g non-T2D 6
g/a non-T2D g/g non-T2D 5
g/g T2D
A g/g T2D B A B
Background g/a non-T2D Increased in
5
g/a T2D 2 4 g/a T2D
Test Set
4
4
Test Set
1
Test Set
3
Test Set g/a UCP3 UCP3 Effects
3 2
2 1
0
Biosynthetic Sources of NEFAs, eCBs and OxLs 2
0 -3 -2 -1 0 1 2 3 1
non-T2D
to[1]
-4 -3 -2 -1 0 1 2
to[1]
to[1]
to[1]
1 -1 0
0.5
0
-2 -3 -2 -1
-1
0 1 2 3 4
T2D
-5 -4 -3 -2 -1 0 1 2 3 -2
-1 -4 -2
-3
0
Loadings1
-2 -3
-6
-3 -4
-4
-8
-4 -5 -5 -0.5 5-LOX LA-polyols n-3 OXL SEA
t[1] t[1]
t[1] t[1]
-1
Fig 4: Scores plots of T2D v non-T2D OPLS-DA models for (A) g/g Fig 6: Scores plots of g/g v g/a UCP3 OPLS-DA models for (A) non-
UCP3 and (B) g/a UCP3 subjects. Sub-group discrimination is greatly T2D and (B) T2D subjects. Models can predict UCP3 genotype better -1.5
improved relative to whole population analysis (see Fig 2). in non-T2D compared to T2D subjects.
Increased
g/a UCP3 VIP Increased in VIP Summarized effects of G304A UCP3:
in T2D g/a UCP3
g/g UCP3 non-T2D
Bioactive lipids involvement in T2D development 1. linoleate di- and tri- hydroxyls may be related to
1
1 T2D increased oxidative stress.
0.8
0.6
4 3 1 6 2 7 5
2. 5-LOX metabolites in non-T2D may signify
0.5
0.4
4 3 1 6 2 7 5
decreased stimulation of insulin secretion
0.2
0 3. steroylethanolamide (SEA) in T2D, an
Loadings 1
Loadings 1
0 endogenous PPARα agonist.
-0.2
n-3 OXL
5-LOX
oxidation
MUFA/Sat.
MAGs
NAEs
Epoxides
LA-polyols
SEA
-0.5 Auto-
Increased in
C18-
MUFA/Sat.
PUFA
5-LOX
NA-AA
MAGs
NAEs
epoxides
Epoxides
n-3 OXL
oxidation
LA-polyols
VLC-
-0.4
Auto-
C18-
C20-
g/a UCP3 T2D x UCP3 Effects
-0.6 0.4
-1 non-T2D
-0.8 0.3
T2D
-1 Param eter Groups
4 3 1 6 2 7 5 0.2
4 3 1 6 2 7 5 -1.5 Param eter Groups
0.1
Methods Fig 5: Bar graph of loadings of T2D v non-T2D OPLS-DA models for Fig 7: Bar graph of loadings of g/g v g/a UCP3 OPLS-DA models
1
Loadings
UCP3 sub-populations. The relative importance of metabolites for T2D for the T2D sub-populations. Relative increases in LA-polyols are 0
Metabolomic Analysis discrimination is influenced by UCP3 polymorphism. associated with g/a UCP3 regardless of T2D status. -0.1
• Fasting plasma MAGs NAEs C18-Epoxides MUFA/Sat. Auto-oxidation
• Selected Reaction Monitoring mode LC-ESI-MS/MS (OxLs and eCBs) -0.2
• Selected Ion Monitoring mode GC-EI-MS (NEFA) Multivariate Linear Parameter Connectivities are Shifted by T2D Status
-0.3
Statistical Analysis Software Varied
Decreased Increased in T2D
• R (http://www.R-project.org)
Fig 8: Undirected graph
A B
• SIMCA-P+ 12.0 (http://www.umetrics.com/simca) -0.4 in T2D
MAGs eCBs and OxL representing multi-
Abbreviations
NEFA (non-esterified free fatty acids); eCBs (endocannabinoids); OxL (oxylipins); T2D (type dimensionally scaled Evidence of UCP3 and T2D interaction:
2 diabetic); UCP3 (uncoupling protein 3); PUFA (polyunsaturated fatty acid); Sat correlations among
(saturated fatty acid); MAG (monoglycerol); NAE (N-acylethanolamide); FA-AA (fatty BMI 1. Compared to g/g carriers, T2D-associated metabolites
acyl amino acid); c18 (18 carbon); c20 (20 carbon); VIP (variable’s importance on parameters in (A) non-
projection); PPAR (peroxisome proliferator-activated receptor); n-3 (omega-3)
T2D (B) T2D subjects.
are:
Data Interpretation Strategy NAEs Graphed edges (i.e. lines in g/a non-T2D subjects
BMI between nodes) indicate
Univariate Tests
(t-tests with FDR)
Megavariate Analysis Empirical Metabolic Networks positive (orange) and in g/a T2D subjects
MAGs negative (blue) bivariate
fasting glucose correlations (α=0.05). Interpretative Summary
PCA OPLS-DA Correlation Analysis
and HbA1c C18- Increases in node
epoxides proximity indicates Targeted lipidomic analysis including NEFA, eCBs,
increases in multivariate and OxLs successfully discriminate both T2D and
NEFA correlation strength.
Feature Selection
Multi-Dimensional
Scaling UCP3 dependent phenotypes.
Ellipses indicate the 95%
confidence interval of Circulating markers of UCP3 activity suggest
fasting glucose lipid classes grouped by
Model Comparison Model Validation
Significance Testing
NEFA impacts on insulin signaling and oxidative stress
and HbA1c biochemical and
structural similarities.
responses.
Biological Interpretation Network Visualization
Non-T2D Parameter Connectivity T2D Parameter Connectivity UCP3 dysfunction reduces markers of T2D
severity in clinical T2D subjects.
This work was supported by USDA-ARS Project 5306-51530-016-00D, NIH-NIDDK R01DK078328-01, T32-GM08799, NIH grants DK-038764, DK-083562, and P01 HL-055782 and the Merit Review program of the Department of Veterans Affairs (W.T.G.). The authors also acknowledge support from the research core facilities of the UAB Center for Clinical and Translational
Science (UL1 RR025777), the UAB Nutrition and Obesity Research Center (P30-DK56336), and the UAB Diabetes Research and Training Center (P60 DK079626).