wells-wood-research / timed-design Goto Github PK
View Code? Open in Web Editor NEWProtein Sequence Design with Deep Learning and Tooling like Monte Carlo Sampling and Analysis
Protein Sequence Design with Deep Learning and Tooling like Monte Carlo Sampling and Analysis
More information [in our docs](https://docs.streamlit.io/library/advanced-features/caching).
2024-01-31 13:24:56.783 `st.cache` is deprecated. Please use one of Streamlit's new caching commands,
`st.cache_data` or `st.cache_resource`. Based on this function's return value
of type `str`, we recommend using `st.cache_data`.
DuplicateWidgetID: There are multiple identical st.selectbox widgets with key='option'.
To fix this, please make sure that the key argument is unique for each st.selectbox you create.
Traceback:
File "/scratch/channels/timed-design/ui.py", line 950, in <module>
main(params)
File "/scratch/channels/timed-design/ui.py", line 904, in main
slice_seq, slice_real, real_metrics = _draw_output_section(
File "/scratch/channels/timed-design/ui.py", line 481, in _draw_output_section
option = st.selectbox(
...
Add ability to return to session
Output of predict.py saves all files but .fasta to directory specified in command line --path_to_output OUTPUT option:
python predict.py --path_to_dataset DATASET --path_to_model MODEL --path_to_output OUTPUT
.fasta files are saved to the current working directory
I would love to be able to specify residues to design. Or full regions. If I have a structure, and I want to be able to (re)design a loop.
Due to a bug in the training of the models, the training frames for the Charge and Polar models had the order:
[C, N, O, Cb, Ca, CHARGE/POLARITY]
Rather than
[C, N, O, Ca, Cb, CHARGE/POLARITY]
Which makes the Charge and Polar models unusable. A very hacky quick-fix of this would involve swapping the channels order at prediction time until we are able to spend time and computation retraining the models.
I propose to modify the function load_batch
at
timed-design/design_utils/utils.py
Line 487 in abc6afa
to include something like this ONLY FOR CHARGE AND POLAR MODELS:
# Extract frame from batch:
for i, (pdb_code, chain_id, residue_id, _) in enumerate(data_point_batch):
# Extract frame:
residue_frame = np.asarray(dataset[pdb_code][chain_id][residue_id][()])
# Check if the frame has the correct shape (final dimension is 6) for swapping:
if residue_frame.ndim == 4 and residue_frame.shape[-1] == 6:
# Swap only the 4th and 5th channels (index 3 and 4)
residue_frame[..., 3], residue_frame[..., 4] = residue_frame[..., 4].copy(), residue_frame[..., 3].copy()
X[i] = residue_frame
# Extract residue label:
y[i] = dataset[pdb_code][chain_id][residue_id].attrs["encoded_residue"]
The charge and polar have a final dimension of 6
Hi @universvm, nice work!
I just want to ask if you have a preferred way to cite the project!
Thanks
May or may not be related to #36
How much compute do you need to run a sequence logo? More than the compute available in the universe.
A frequent residue numbering alongside the designed sequence. Perhaps every 10 or 5 residues have the residue number. It would make studying specific residue changes easier.
I was looking at specific residues which interact with a cofactor to see if TIMED offered alternative residues, but this involved counting along the sequence until we found the residues, a numbering would help:)
flat_dataset_map, prediction_matrix, rotamers_categories=None
AssertionError: Prediction Matrix file TIMED1qys.csv does not exist
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.