GitHub - FowlerLab/Envision2017: We present Envision, an accurate predictor of protein variant molecular effect, trained using large-scale experimental mutagenesis data. All data and software in this study are freely available. The training data set and all code used to train the models and generate the figures presented in this manuscript are available here. Envision predictions, along with feature annotations, are available at https://envision.gs.washington.edu/.

FowlerLab / Envision2017 Public

Notifications You must be signed in to change notification settings
Fork 1
Star 16

We present Envision, an accurate predictor of protein variant molecular effect, trained using large-scale experimental mutagenesis data. All data and software in this study are freely available. The training data set and all code used to train the models and generate the figures presented in this manuscript are available here. Envision predictio…

16 stars 1 fork Branches Tags Activity

Star

Notifications

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
data		data
LOPOTuneTrain.ipynb		LOPOTuneTrain.ipynb
LOPO_10xCV.ipynb		LOPO_10xCV.ipynb
LOPO_predict_missingFeatureMuts.ipynb		LOPO_predict_missingFeatureMuts.ipynb
LOPO_unnormalized.ipynb		LOPO_unnormalized.ipynb
README.txt		README.txt
downSamplingAnalysis.ipynb		downSamplingAnalysis.ipynb
envisionTuneTrainPredict.ipynb		envisionTuneTrainPredict.ipynb
envision_figure_code.Rmd		envision_figure_code.Rmd
envision_figure_code.html		envision_figure_code.html
singleProteinModels.ipynb		singleProteinModels.ipynb

Repository files navigation

Our code is separated into five Jupyter Notebook files (.ipynb) and one
R Markown file.

The Jupyter Notebooks contain the following:
------------------------------------------------------------------------
+ singleProteinModels.ipynb -- code for tuning hyperparameters and
training models using the 8 protein data sets individually.

+ envisionTuneTrainPredict.ipynb -- code to tune hyperparameters and
train Envision with all eight data sets

+ LOPOTuneTrain.ipynb -- train each leave-one-protein-out
(LOPO) model to predict the protein data set not used in training.

+ LOPO_10xCV.ipynb -- tune using tenfold cross-validation, train each leave-one-protein-out
(LOPO) model to predict the protein data set not used in training.

+ LOPO_predict_missingFeatureMuts.ipynb -- use each leave-one-protein-out
(LOPO) model to predict the protein data set not used in training with missing features.

+ LOPO_unnormalized.ipynb -- train each leave-one-protein-out
(LOPO) model with unnormalized data and then predict protein data sets not used in training.

+ downSamplingAnalysis.ipynb -- code to sample 6, 4,and 2 proteins
as training data for model training

+ Clinvar_analysis.ipynb -- use Envision to predict Clinvar mutations
_______________________________________________________________________

The R Markdown contains the following:
---------------------------------------------------------------------

+ envision_figure_code.Rmd -- code for generating manuscript figures.
---------------------------------------------------------------------

Notes:
- All necessary data files can be found in /data directory.

- Graphlab and Python dependencies (e.g. Numpy) are required to
successfully run all .ipynb code.

- All code will be deposited in a public GitHub repository upon publication