Organisation of the dissertation

As it is a fruit of an interdisciplinary work, I decided to introduce the topic from two perspectives: describe the biological and biomedical dimension of the topic (see Chapter 1), as well as, the mathematical dimension of the problem of separation of sources in complex mixtures (see Chapter 2). I hope, it will make the subject of my thesis easy to understand also for non-biologists or non-mathematicians. In the results part, I introduce a study of ICA applied to transcriptomes (Chapter 3). I also apply ICA-based deconvolution to Breast cancer transcriptomes to prove its reproducibility Chapter 4. I compare the reproducibility of blind source separation methods NMF and ICA (see Chapter 5). Then I introduce the DeconICA R package (see Chapter 6 ) and finally present results of an application of DeconICA and other tools to 118 transcriptomic datasets (see Chapter 7). The second part of the results is dedicated to my work on cell type heterogeneity (see Chapter 8). The manuscript finishes with Chapter 9 and Chapter 10 that contain discussion, conclusions, and perspectives. In annexes, you can find publications to which I contributed during my doctorate that are not strictly linked with the topic of this thesis. In the end, I included a glossary of useful terms.

INTRODUCTION

  • Chapter 1: introduction to cancer biology and immunity, challenges in cancer immunotherapies and cancer immune phenotyping as well as data sources most commonly used to face the topic.

  • Chapter 2: introduction to a problem of mixed sources in biological samples, an overview of blind source separation methods and supervised deconvolution methods, with focus on those applied to bulk transcriptome to uncover and quantify immune compartments

RESULTS

  • Chapter 3: Most Reproducible Transcriptome Dimension (MSTD)
  • Chapter 4: application of ICA-based deconvolution to six breast transcriptomes
  • Chapter 5: comparison of reproducibility of NMF and ICA methods
  • Chapter 6: DeconICA R package
  • Chapter 7: application of DeconICA R package and other tools to analyze >100 transcriptome datasets of bulk cancer transcriptomes
  • Chapter 8: study of immune cell types heterogeneity in tumor microenvironment using the innate immune map and scRNA-seq data

DISCUSSION

ANNEXES

  • Other publications:
    • Adjustment of dendritic cells to the breast-cancer microenvironment is subset specific
    • The inconvenience of data of convenience: computational research beyond post-mortem analyses
  • DeconICA R package documentation:
    • Vignette 1: Introduction to deconICA
    • Vignette 2: Running fastICA with icasso stabilization
    • Manual
  • Scientific CV (including a list of attended conferences and publications)

GLOSSARY