Annexes
1 DeconICA documentation
Tutorials and manual are available as a part of R package documentation (Vignettes and Reference Manual) at https://github.com/UrszulaCzerwinska/DeconICA.
1.1 Introduction to deconICA
See the online tutorial Introduction to deconICA at: https://urszulaczerwinska.github.io/DeconICA/DeconICA_introduction.html
1.2 Running fastICA with icasso stabilisation
See the online tutorial Running fastICA with icasso stabilisation at: https://urszulaczerwinska.github.io/DeconICA/Icasso.html
1.3 Reference manual
The formal documentation of the R package automatically generated. Describes all functions and data with examples included in the package.
The manual is available from the R package and online: https://github.com/UrszulaCzerwinska/DeconICA/blob/master/inst/manual/DeconICA.pdf
2 Publications and conferences
In this section I included publications with my minor contribution or not related to the thesis topic. At the end of the section, there is my CV with listed publicatons, conferences and selected graduated courses.
2.1 Adjustment of dendritic cells to the breast-cancer microenvironment is subset-specific
Paula Michea\(^\star\), Floriane Noël\(^\star\), Eve Zakine, Urszula Czerwinska, Philemon Sirven, Omar Abouzid, Christel Goudot, Alix Scholer-Dahirel, Anne Vincent-Salomon, Fabien Reyal, Sebastian Amigorena, Maude Guillot-Delost, Elodie Segura, and Vassili Soumelis
\(^\star\) \(^{_{contributed}}\) \(^{_{equally}}\)
Published in Nature Immunology on 16th July 2018
This project developed by Michea et al. originated in Vassili Soumelis group with interesting quality bulk RNa-seq data on pDC cells subsets in breast cancer.
On my side, I worked on an alternative to DGE (presented in this publication) approach aiming to verify if the pDC subsets can be discovered in an unsupervised manner from the data. As there were a number of samples for each subset available, I used ICA to decompose each subset. With the ICA components I created a correlation network of common and subset-specific signals.
On the other hand, I computed module activity scores with ROMA software (Martignetti et al. 2016) of each samples using a wide collection of pathways and then used hierarchical clustering to order the samples.
In my analysis, some subsets were clearly separated (MMAC, BDCA1pDC) and some not (BDCA1nDC and CD14pDC).
A strategical decision was taken to not to include my part of work in the main storyline. I actively participated in article writing and the review process.
Article available online: https://www.nature.com/articles/s41590-018-0145-8
2.2 The inconvenience of data of convenience: computational research beyond post-mortem analyses
Chloé-Agathe Azencott, Tero Aittokallio, Sushmita Roy, DREAM Idea Challenge Consortium, Thea Norman, Stephen Friend, Gustavo Stolovitzky & Anna Goldenberg
DREAM Idea Challenge Consortium:
Ankit Agrawal, Tero Aittokallio, Chloé-Agathe Azencott, Emmanuel Barillot, Nikolai Bessonov, Deborah Chasman, Urszula Czerwinska, Alireza Fotuhi Siahpirani, Stephen Friend, Anna Goldenberg, Jan Greenberg, Manuel Huber, Samuel Kaski, Christoph Kurz, Marsha Mailick, Michael Merzenich, Nadya Morozova, Arezoo Movaghar, Mor Nahum, Torbjörn E M Nordling, Thea Norman, Robert Penner, Sushmita Roy, Krishanu Saha, Asif Salim, Siamak Sorooshyari, Vassili Soumelis, Alit Stark-Inbar, Audra Sterling, Gustavo Stolovitzky, S S Shiju, Jing Tang, Alen Tosenberger, Thomas Van Vieet, Krister Wennerberg & Andrey Zinovyev
Published in Nature Methods on 29 September 2017
One of burning problems of computational scientists is the fact that the data ideal to verify some hypothesis born from theoretical work or simulations do not exist. Idea Dream Challenge was a call for projects that would describe the ideal data for a proposed model. Winning project would obtain money to gather necessary data. All projects would participate in matching board that aimed to expose interesting theoretical work with experimental scientist that may have or produce necessary data.
I proposed a project, together with my thesis supervised Andrei Zinovyev and Vassili Soumelis. We proposed three independent datasets that could be used to study TME.
- A single cell data from tumor transcriptomes filling requirements of minimal number of cells of each type to facilitate the statistical analysis.
- A bulk transcriptome data of systematically co-culture immune-related cells of different types together controlling their proportions with sufficient number of combinations (at least several hundreds) of different cell type proportions in order to study cell-cell interactions. This data would contain 1) individual transcriptomic profiles of pure cell cultures (few tens, containing the replicas) and 2) transcriptomic profiles of controlled mixtures of cell cultures (if possible, containing combinations of many cell types).
- A benchmark dataset for deconvolution methods: bulk transcriptomic profiles of tumoral samples coupled with carefully quantified proportions of the immune-related cells of different types and the tumoral cellularity .
In the project description we presented as well the ICA model of deconvolution and our preliminary results.
Our project was selected in the first but not the second round of the review process.
All participant of the Idea Dream Challenge co-authored the correspondence to Nature Mathods as the DREAM Idea Challenge Consortium.
Article available online: https://www.nature.com/articles/nmeth.4457
2.3 CV: publications, conferences, courses
References
Martignetti, Loredana, Laurence Calzone, Eric Bonnet, Emmanuel Barillot, and Andrei Zinovyev. 2016. “ROMA: Representation and Quantification of Module Activity from Target Expression Data.” Front. Genet. 7: 18. doi:10.3389/fgene.2016.00018.