Skip to content

Latest commit

 

History

History

examples

README.md

SeqDesign Examples

Downloading data

Example data

./download_example_data.sh

This script will download the following files for training, mutation prediction, and sequence generation:

  • datasets/sequences/BLAT_ECOLX_1_b0.5_lc_weights.fa
  • datasets/nanobodies/Manglik_filt_seq_id80_id90.fa
  • datasets/nanobodies/Manglik_labelled_nanobodies.txt
  • calc_logprobs/input/BLAT_ECOLX_r24-286_Ranganathan2015.fa
  • sess/BLAT_ECOLX_v2_channels-48_rseed-11_19Aug16_0626PM.ckpt-250000*
  • sess/nanobody.ckpt-250000*

Training sequences

./download_sequences.sh

This script will download all training sequences from the paper to datasets/sequences/ and datasets/nanobodies/

Effect predictions

./download_effect_predictions.sh

This script will download all mutation effect predictions from the paper to calc_logprobs/output/

Generated sequences

./download_generated_nanobodies.sh

This script will download the designed nanobody library to generated/

Running examples

Training the model

./demo_train.sh

This script will run 100 training iterations on the β-Lactamase sequence dataset (the full model runs 250,000 iterations). The final model checkpoint will appear as three files in sess/BLAT_ECOLX_elu_channels-48_rseed-11_<timestamp>.ckpt-100*

On an AWS p2.xlarge instance, this demonstration took 2 minutes.

Predicting mutation effects

./demo_calc_logprobs.sh

This script will use the pretrained model weights in sess/BLAT_ECOLX_v2_channels-48_rseed-11_19Aug16_0626PM.ckpt-250000* to make mutation effect predictions for the β-Lactamase mutational scan from Stiffler et al., Cell, 2015.

The final predictions are the average of 10 predictions (500 are used in the full test). These predictions will appear in calc_logprobs/output/demo_BLAT_ECOLX_r24-286_Ranganathan2015_rseed-11_channels-48_dropoutp-0.5.csv

On an AWS p2.xlarge instance, this demonstration took 3.5 minutes.

Generating nanobody libraries

./demo_generate.sh
./demo_generate_fast.sh

This will generate nanobody CDR3 and FRA4 sequences given a preceding VH sequence. The full nanobody sequences will be output in generated/nanobody.ckpt-250000_temp-1.0_rseed-42.fa

On an AWS p2.xlarge instance, demo_generate.sh took 2.5 minutes and demo_generate_fast.sh took 30 seconds.