To use our original or adapted codes, please cite our work https://doi.org/10.1101/2020.02.13.943688 as [1], see reference at the end of this document.
The MATLAB code present in this directory has been written by Stefano Magni and Rucha Sawlekar, with contributions from Laurent Mombaerts and Zuogong Yue. It is based on scientific (computational and experimental) work performed by Stefano Magni, Rucha Sawlekar, Jorge Goncalves, Laurent Mombaerts, Zuogong Yue, Feng He, Christophe Capelle, Ye Yuan, Alexandre Baron and Ni Zeng. This work can be found as a preprint in the manuscript [1] which can be downloaded from BiorXiv https://doi.org/10.1101/2020.02.13.943688 (currently submitted for journal publciation) and titled "Causal dynamical modelling predicts novel regulatory genes of FOXP3 in human regulatory T cells".
This code represents the implementation of our method for dynamically modelling gene regulations between genes starting from time series microarray data measuring gene expression. In particular, this is an application of our modelling strategy to the inference of regulators of FOXP3 in human T regulatory cells. This code was used to generate all the results (plus additional ones) described in the above mentioned manuscript, in particular the ranking of genes as potential regulators of FOXP3. We suggest to read our manuscript above in order to understand the scientific scope of this code. Note that these scripts are meant to be run on the time-series data generated by He et al. (2012), https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3531908/. The raw data from that study can be found at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE11292 under "Samples".
This repository contains different types of folders.
- Folders whose names start with Preliminary: they contain scripts which perform preliminary steps of pre-processing of the raw data mentioned above, to normalize them (gcrma) and load them into MATLAB in a format suitable for analysis. these scripts should be run once on the raw data, and they are necessary for all the following scripts to run.
- Folders whose names start with MAJOR: they contain all the scripts which where used to filter the above mentioned data (MAJOR_1), apply the All2one method without delay (MAJOR_2), and with delay (MAJOR_3), which means to generate all the results of our manuscript above. In addition, there ate further analysis on Teff cells (MAJOR_4).
- Folders whose names start with RESULTS: these folders contain results of the ranking performed by the previous scripts.
- Folders whose names start with Minor: these folders contain various additional analysis which were performed and are not covered in the manuscript. Certain investigated further aspects of the system under study, but didn't contribute to the focus of this manuscript. Others do not allow to extract any useful information.
- Folder named Manuscript: this folder contains the manuscript itself.
- Folder Named Dynamics_of_4TargetGenes_Tregs: this folder contains plots of the time series data of FOXP3, and of three other genes of potential interest for Tregs.
To run our code:
- Download the full content of the directory containing this README file.
- Make sure you have installed MATLAB (we developed the code under versions 2016a, 2016b and 2017a, so we suggest these or subsequent versions).
- Open MATLAB.
- Select the script corresponding to the analysis you are interested in running, and run it.
Our method contained in this repository has the main goal of scanning genome-wide time-series microarray from human T regulatory cells data and rank genes as potential regulators of FOXP3. The dynamical modelling technique used here employs first order linear-time invarian (LTI) O.D.E. models, and it is based on previous work from members of our group, primarily Laurent Mombaerts. The codes from that work can be found e.g. in https://github.com/Lmombaerts/DyDE, and the study is described in [2]. A comparison of the performances of the DyDE technique w.r.t. other methods for inference of gene regulatory networks can be found in [3]. The DyDE technique has also been applied to the study of barley in [4].
[1] Sawlekar, R., Magni, S., Capelle, C. , Baron, A., Zeng, Ni, Mombaerts, L., Yue, Z., Yuan, Y., He, F.Q. and Gonçalves, J. (2020) Causal dynamical modelling predicts novel regulatory genes of FOXP3 in human regulatory T cells. BioRxiv 2020.02.13.943688; doi: https://doi.org/10.1101/2020.02.13.943688
[2] Mombaerts, L., Carignano, A., Robertson, F. C., Hearn, T. J., Junyang, J., Hayden, D., ... & Yuan, Y. (2019). Dynamical differential expression (DyDE) reveals the period control mechanisms of the Arabidopsis circadian oscillator. PLoS computational biology, 15(1), e1006674.
[3] Mombaerts, L., Aalto, A., Markdahl, J., & Gonçalves, J. (2019). A multifactorial evaluation framework for gene regulatory network reconstruction. IFAC-PapersOnLine, 52(26), 262-268.
[4] Müller, L., Mombaerts, L., Pankin, A., Davis, S. J., Webb, A. A., Goncalves, J., & von Korff, M. (2019). Differential effects of day-night cues and the circadian clock on the barley transcriptome. bioRxiv, 840322.