View
25
Download
0
Category
Preview:
Citation preview
JOBIM NANTES – 04/07/19
Etienne CAMENEN
Ivan MOSZER, François-Xavier LEJEUNE & Arthur TENENHAUSiCONICS core facility, Institut du Cerveau et de la Moelle épinière, Hôpital Pitié Salpêtrière, Paris
Multimodal data analysis:a Shiny & Galaxy approach
??
CONTEXT
JOBIM, Nantes, July 4th, 2019
Multi-modal
1
x1
v1 v2
x1 0.2 0.3
x2 0.7 NAx2
z1 z2
x1 0 0.1
x2 0.5 -1
x3 1 NA...
... ... ...
Modalities(blocks)
y1 y2
x1 0.2 0.3
x2 0.7 NA
JOBIM, Nantes, July 4th, 2019
Multi-way
1
v1 v2
x1 0.2 0.3
x2 0.7 NA
v1 v2
x1 0.2 0.3
x2 0.7 NA
v1 v2
x1 0.2 0.3
x2 0.7 NA
...
...
...
x1
x2
...
JOBIM, Nantes, July 4th, 2019
Multi-group
1
v1 v2
x1 0.2 0.3
x2 0.7 NA
v1 v2
xa 0.2 0.3
xb 0.7 NA
xa
xb
...
...
x1
x2
...
...
JOBIM, Nantes, July 4th, 2019
Multi-modal
1
x1
v1 v2
x1 0.2 0.3
x2 0.7 NAx2
z1 z2
x1 0 0.1
x2 0.5 -1
x3 1 NA...
... ... ...
Modalities(blocks)
y1 y2
x1 0.2 0.3
x2 0.7 NA
JOBIM, Nantes, July 4th, 2019
Objectives
❖ IntegrParkinson: developing graphical interfaces for RGCCA (Regularized Generalized Canonical Correlation Analysis) to interactively analyze multimodal & high-dimensional data➢ Galaxy & Shiny➢ Targeted to biologists
❖ Use cases:➢ Parkinson Disease (Nucleipark; S. Lehéricy & M. Vidailhet)➢ ORL cancers (PREDICTOR; P. Garteiser & B. Van Beers)➢ Portal hypertension (HTP; D. Visvikis & M. Hatt)
2
JOBIM, Nantes, July 4th, 2019
Parkinson Disease
Nucleipark
P = ~2500
RNA-SeqP = ~20.000
N = 52
N = 25
RMI 3TP = 5
P = 5
OOH P = ~400
Scores: Motor / Non-motor
MRI Functional networks
Monocyte
LC-MSPlasma
3
DEVELOPMENT
JOBIM, Nantes, July 4th, 2019
Three delivrables
RGCCA
Package Executable
4
JOBIM, Nantes, July 4th, 2019
User-friendly
Before After
...
1 argument
5
JOBIM, Nantes, July 4th, 2019
Three delivrables
RGCCA
Package Executable Interface
compatibility
6
EXECUTABLE
JOBIM, Nantes, July 4th, 2019
INPUT(S).csv, .tsv, .xls
Executable
7
y1
x1 0
x2 1
v1 v2 v3
x1 0.2 0.3 3.1
x2 0.7 NA 7.4
z1 z2
x1 14 1.8
x2 50 -2
x3 28 1.1
JOBIM, Nantes, July 4th, 2019
INPUT(S)RGCCA.csv, .tsv, .xls
Executable
? ?
7
- correlation inter-modalities
- understand individuals relations
- variable reduction
y1
x1 0
x2 1
v1 v2 v3
x1 0.2 0.3 3.1
x2 0.7 NA 7.4
z1 z2
x1 14 1.8
x2 50 -2
x3 28 1.1
JOBIM, Nantes, July 4th, 2019
INPUT(S) OUTPUTRGCCA.csv, .tsv, .xls .png, .jpeg,
.pdf, .tiff
Executable
? ?
7
- correlation inter-modalities
- understand individuals relations
- variable reduction
y1
x1 0
x2 1
v1 v2 v3
x1 0.2 0.3 3.1
x2 0.7 NA 7.4
z1 z2
x1 14 1.8
x2 50 -2
x3 28 1.1
JOBIM, Nantes, July 4th, 2019
INPUT(S) OUTPUTRGCCA.csv, .tsv, .xls .png, .jpeg,
.pdf, .tiff
Executable
? ?
7
- correlation inter-modalities
- understand individuals relations
- variable reduction
y1
x1 0
x2 1
v1 v2 v3
x1 0.2 0.3 3.1
x2 0.7 NA 7.4
z1 z2
x1 14 1.8
x2 50 -2
x3 28 1.1
INTERFACES
JOBIM, Nantes, July 4th, 2019
Connexion with other tools
Dynamic & interactive
Comparaison
Large community
8
JOBIM, Nantes, July 4th, 2019
❖ Public web platform : – tools database to analyze omics data
9
JOBIM, Nantes, July 4th, 2019
❖ Public web platform : – tools database to analyze omics data– ad hoc pipeline creation
Interoperability (e.g., enrichment)
9
JOBIM, Nantes, July 4th, 2019
❖ Interface – web application,
launched from Rstudio
– dynamic parameter settings
❖ Graphical interface R:
10
PERSPECTIVES
JOBIM, Nantes, July 4th, 2019
❖ Statistics:– Missing values (C. PELTIER, poster n°53), cross-validation
(A. BOURRELIER), temporality (multiway; A. GLOAGUEN), etc.
❖ Informatics:– Drag-n-drop: blocks + connection– Dynamic visualization
❖ Data analysis:– Support partners to identify potential biomarkers
Perspectives
11
ACKNOWLEDGEMENTS
iCONICS team:- Ivan MOSZER- Arthur TENENHAUS- François-Xavier LEJEUNE- Caroline PELTIER- Arnaud GLOAGUEN
IFB partners
Station de Roscoff- Gildas LE CORGUILLE- Romain DALLET
Recommended