pysero
pysero enables serological measurements with multiplexed and standard ELISA assays.
The project automates estimation of antibody titers from data collected with ELISA assays performed with antigen-arrays and single antigens.
The immediate goal is to enable specific, sensitive, and quantitative serological surveys for COVID-19.
Installation
On a typical Windows, Mac, or Linux computer:
- Create a conda environment:
conda create --name pysero python=3.7 - Activate conda environment:
conda activate pysero - Pip install dependencies:
pip install -r requirements.txt - Add the package to PYTHONPATH. Inside the package directory (...\serology-COVID19), do:
export PYTHONPATH=$PYTHONPATH:$(pwd)
For installation notes for Jetson Nano, see these notes.
Usage
usage: pysero.py [-h] (-e | -a) -i INPUT -o OUTPUT
[-wf {well_segmentation,well_crop,array_interp,array_fit}]
[-d] [-r] [-m METADATA]
optional arguments:
-h, --help show this help message and exit
-e, --extract_od Segment spots and compute ODs
-a, --analyze_od Generate OD analysis plots
-i INPUT, --input INPUT
Input directory path
-o OUTPUT, --output OUTPUT
Output directory path, where a timestamped subdir will
be generated. In case of rerun, give path to
timestamped run directory
-wf {well_segmentation,well_crop,array_interp,array_fit}, --workflow {well_segmentation,well_crop,array_interp,array_fit}
Workflow to automatically identify and extract
intensities from experiment. 'Well' experiments are
for standard ELISA. 'Array' experiments are for ELISA
assays using antigen arrays printed with Scienion
Array Printer Default: array_fit
-d, --debug Write debug plots of well and spots. Default: False
-r, --rerun Rerun wells listed in 'rerun_wells sheets of metadata
file. Default: False
-m METADATA, --metadata METADATA
specify the file name for the experiment metadata.
Assumed to be in the same directory as images.
Default: 'pysero_output_data_metadata.xlsx'
Extract OD from antigen array images
python pysero.py -e -i <input> -o <output> -m <METADATA> will take metadata for antigen array and images as input, and output optical densities for each antigen. The optical densities are stored in an excel file at the following path: <output>/pysero_<input>_<year><month><day>_<hour><min>/median_ODs.xlsx
If rerunning some of the wells, the input metadata file needs to contain a sheet named 'rerun_wells' with a column named 'well_names' listing wells that will be rerun.
This workflow describes the steps in the extraction of optical density.
Generate OD analysis plots
python pysero.py -a -i <input> -o <output> -m <METADATA> will read pysero or scienion spot fitting outputs and generates analysis plots for each single antigen. 3 types of plots are supported for now (ROC, categorical, standard curves). The example xlsx config file can be found in \example folder in the repo.
An '-l' flag can be added to load the saved report from previous run to speed up loading.
Train a classifier using information from multiple antigens
One could train a machine learning classifier using ODs from multiple antigens to potentially improve the classification accuracy for sero-positive or sero-negative. The following script demonstrates how to do this with xgboost tree classifiers. python -m interpretation.train_classifier
Equipment list
The project aims to implement serological analysis for several antigen multiplexing approaches.
It currently supports:
- classical ELISA.
- antigen arrays printed with Scienion.
It can be extended to support:
The antigen-arrays can be imaged with:
- any transmission microscope with motorized XY stage.
- turn-key plate imagers, e.g., SciReader CL2.
- Squid - a variant of Octopi platform from Prakash Lab.
The project will also have tools for intersecting data from different assays for estimation of concentrations, determining level of cross-reactivity, ...
Validation
Current code is validated for analysis of anigen arrays imaged with Scienion Reader and is being refined for antigen arrays imaged with motorized XY microscope and Squid.
Contributions
We welcome bug reports, feature requests, and contributions to the code. Please see this page for most fruitful ways to contribute.
0 Apr 22, 2022
18 Dec 21, 2022
907 Jan 4, 2023
2 Aug 30, 2022
2 Jan 13, 2022
3 Dec 8, 2022