This repository contains the code used for the paper "Flow cytometric single-cell identification of populations in synthetic bacterial communities".
InSilicoFlow depends on the following packages:
- Python 3 (version 3.4.5)
- numpy (version 1.11.1)
- scikit-learn (version 0.18)
- pandas (version 0.18.1)
Script used to perform the first part of the analysis explained in the paper. It mainly performs the following steps:
- Choose two or m bacterial populations of interest.
- Aggregate data coming from replicates.
- Aggregate data coming from the populations of interest (creating the so-called in silico community).
- Use 70% of the community to train a classifier (e.g., LDA or Random Forests).
- Use the other 30% to evaluate the performance of a classifier (expressed in for example the accuracy, AUC, ...).
Script used to retrieve the composition of a synthetic bacterial community. It mainly performs the following steps:
- Create in silico community, representing the synthetic community of interest.
- Train classifier on full community.
- Evaluate performance by retrieving community composition for various in silico/in vitro communities.
Our data can be found in .csv-format on this repo. Additionally, our data is made available in .fcs-format on the flowRepository, using the identifiers FR-FCM-ZZSH (axenic cultures) and FR-FCM-ZZSG. It has been preprocessed following the robust digital gating strategy by Prest et al. (2013).
If one wants to start from the raw .fcs-files, it can be transformed to .csv-format using fcstocsv.py; therefore the data first needs to be transformed to a .csv-format; fcstocsv.py takes care of that. It makes use of fcsparser.