aae4seq

Set of tools to encode protein sequences using an Adversarial AutoEncoder and explore the corresponding latent space

installation

Create a conda environment and activate it

conda create --name aae python=3.6
conda activate aae

(for analyses, install ete3)

conda install -c etetoolkit ete3 ete_toolchain 

Install required packages

pip install -r requirements.txt

export src path to PYTHONPATH

export PYTHONPATH=$PWD:$PYTHONPATH

Train a model

Train the example model

python src/aae_model.py -i data/Sulfatases/fasta/hmmsearch_gc90_gs25.fasta -o log/ 

Interpolation

Perform interpolation between sulfatase proteins with pretrained model

python src/interpolate_space.py -i data/fasta/hmmsearch_gc90_gs25.fasta.gz -c data/fasta/hmmsearch_gc90_gs25_encoded.npz -d data/sulfatase_decoder_weights.h5 -q 0 -t 10000 -o interpolated_seq1_seq10000.fasta -s 50