Runnable code to reproduce results presented

Folder contents

Jupyter notebooks that reproduce the paper results:

Figure 2 notebook
Figure 3 notebook (also includes SI Tables)
Figure 4 notebook

Data for reproducing results:

AAV2 capsid sequences (allseqs_20191230.csv.zip)
- Capsid packaging assay results for 297k capsid sequences
- Model training and validation datasets
- Baseline additive and random datasets
- Model-designed sequences
- Model-selected sequences
Subsampled model-designed sequence partitions (subsampled.tar.gz)
- Used to normalize sequence subset sizes prior to clustering
Clustering of all (viable + non-viable) model-designed sequences (clusters.tar.gz)
Clustering of viable-only model-designed sequences (viable_clusters.tar.gz)

Setup Python environment

Install Anaconda distribution of Python

https://www.anaconda.com/products/individual Python 2.7 64-Bit Graphical Installer (637 MB)

Note: instructions have been tested on Mac OS 10.14; Anaconda installers are also available for Windows and Linux.

Install Python dependencies

$ conda install --file requirements.txt

Run notebooks

Start Jupyter notebook server in code directory

$ cd path/to/this/folder $ jupyter notebook .

Navigate your web browser to the URL printed in the console by Jupyter (typically http://localhost:8888).