The repository contains a Python package and Jupyter notebooks used for analysis of genotype-phenotype landscape data. As input, gsf_ims_fitness takes several files produced by software in the bartender-1.1, NISTBartender, and engineering-bio-lacI-landscape repositories (see below). It produces output data in the form of a Pandas DataFrame that can be exported in various formats for additional analysis, if necessary.

gsf_ims_fitness was written in Python3, version 3.11.4

It was written with the following python packages:

• matplotlib

• numpy

• pandas

• scipy

• cmdstanpy

• seaborn

• biopython

• scikit-learn

• palettable

• cmocean

• ipython

• pyyaml

To install the correct package versions, use one of the .yml files included in this repository. gsf_ims_env.yml is from a Windows installation. gsf_ims_env.AWS.yml is from a Linux installation (on an AWS EC2 instance). They are very similar, but they were set up independently, so we reccomend using the environment file spoecific to the computer OS.

After installing conda (miniconda reccomended: https://docs.conda.io/en/latest/miniconda.html), create the gsf_ims environment with:

Windows:

conda env create -f gsf_ims_env.yml

Linux:

conda env create -f gsf_ims_env.AWS.yml

Then install the gsf_ims_fitness package from source:

For Windows:

cd gsf_ims_fitness
pip install -e .

For Linux:

cd gsf_ims_fitness
python3 -m pip install -e .

gsf_ims_fitness requires the following input data files:

• Output from the NISTBartender software: "output_file_label.trimmed_sorted_counts.csv"

• Output from bartender-1.1 software (run as stand-alone or automatically from NISTBartender): "output_file_label_forward_cluster.csv", "output_file_label_forward_barcode.csv", "output_file_label_reverse_cluster.csv", and "output_file_label_reverse_barcode.csv".

• Output from the engineering-bio-lacI-landscape long-read sequencing data analysis pipeline (one file for each plasmid region):

  • "barcode_1.tsv.gz"

  • "barcode_2.tsv.gz"

  • "empty_1.tsv.gz"

  • "empty_2.tsv.gz"

  • "empty_3.tsv.gz"

  • "empty_4.tsv.gz"

  • "insulator.tsv.gz"

  • "KAN.tsv.gz"

  • "lacI.tsv.gz"

  • "Ori.tsv.gz"

  • "primers.tsv.gz"

  • "tetA.tsv.gz"

  • "YFP.tsv.gz"

After generating the required input files listed above, run the following Jupyter notebooks (in order):

This notebook initializes the Pandas Dataframe that will contain the genotype-phenotype landscape data. It also has a step to mark actual chimera reads (dual barcode pairs that are improperly matched) for later removal. Finally, it runs the least-squares fitting routine to covert from barcode read counts to fitness for each sample and runs the Bayesian inference methods to estimate the dose-response curves from the fitness results. The Stan code for the Bayesian inference models is located in the "\gsf_ims_fitness\Stan models" folder in this repository.

This notebook adds columns to the DataFrame to indicate the number of points of agreement between the Hill model and GP model fits, the LacI amino acid sequence for each variant, and the mutation codes indicating the amino acid changes relative to the wild-type sequence (e.g. "A106D")

This notebook uses the output from from bartender-1.1 and engineering-bio-lacI-landscape to add the lacI CDS sequences to the DataFrame.

The following notebooks use the output from bartender-1.1 and engineering-bio-lacI-landscape to add the sequence information for other plasmid regions to the DataFrame:

"PacBio YFP.ipynb", "PacBio tetA.ipynb", "PacBio empty_1.ipynb", "PacBio empty_2.ipynb", "PacBio empty_3.ipynb", "PacBio insulator.ipynb", "PacBio KAN.ipynb", "PacBio Ori.ipynb".

After running the Jupyter notebooks listed above the Pandas DataFrame referenced by the BarSeqFitnessFrame.barcode variable will contain a row for each LacI variant in the library and the following columns of data for each variant: