Update README.md

Author: gwirn

example/README.md

@@ -7,31 +7,58 @@ For a detailed description of the file structure please take a look at [this REA
 **This example assumes it is run from within the base directory of this repository (`image-dms`).**
 
 ## Pre-training dataset creation (optional)
-For a detailed description of the used parameters run `python3 d4_pt_dataset.py`
+This will create a dataset containing pseudo-scores in the `out_path` end will be similar to the output shown in  `gb1None.tsv`.
 
-`python3 d4_pt_dataset.py --p_name gb1 --pdb_file example/gb1.pdb --algn_path example/gb1_1000_experimental.clustal --p_data --p_firstind 0 --p_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --out_path example/ --name_var variant --name_nmut num_mutations --name_score score`
+```sh
+python3 d4_pt_dataset.py --p_name gb1 --pdb_file example/gb1.pdb --algn_path example/gb1_1000_experimental.clustal --p_data --out_path example/ --p_firstind 0 --p_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --name_var variant --name_nmut num_mutations --name_score score
+```
+
+For a detailed description of the used parameters run 
+```sh
+python3 d4_pt_dataset.py -h
+```
 
 ## Pre-training a network (optional)
-For a detailed description of the used parameters run `python3 d4_cmd_driver.py`
+This will pre-train a model on the pseudo-scores in `gb1None.tsv`, store a model with the best weights based on the validation statistics and one from the end of the training which has the suffix `_end` in `result_files/saved_models/gb1None_DD_MM_YYYY_HHMMSS/`, will store the used parameters in `result_files/log_file.csv` and the training statistics in `result_files/results.csv`.
+It can be used in the next step as a starting point for training a network on experimentally determined data.
 
-`python3 d4_cmd_driver.py --query_name gb1 --alignment_file example/gb1_1000_experimental.clustal --tsv_filepath example/gb1None.tsv --pdb_filepath example/gb1.pdb --number_mutations num_mutations --variants variant --score score --wt_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --first_ind 0 --training_epochs 100 --split0 0.7 --split1 0.15 --split2 0.05 --save_model `
+```sh
+python3 d4_cmd_driver.py --query_name gb1 --alignment_file example/gb1_1000_experimental.clustal --tsv_filepath example/gb1None.tsv --pdb_filepath example/gb1.pdb --number_mutations num_mutations --variants variant --score score --wt_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --first_ind 0 --training_epochs 100 --split0 0.7 --split1 0.15 --split2 0.05 --save_model
+```
 
-This will save the network trained on the pre-training dataset created in the previous step. It can be used in the next step as a starting point for training a network on experimentally determined data.
-The saved network can be found in `result_files/saved_models/gb1None_DD_MM_YYYY_HHMMSS/` where the time stamp will depend on the time the training was started.
+For a detailed description of the used parameters run 
+```sh
+python3 d4_cmd_driver.py -h
+```
 
 ## Training a network with experimental data
 In order to train a network on experimentally determined data use one of the methods mentioned below. 
 
-This will train a network on a 0.7-0.15-0.05 training-validation-test split and save the trained model in `result_files/saved_models/gb1None_DD_MM_YYYY_HHMMSS/`
+This will train a network on a 0.7-0.15-0.05 training-validation-test split and save the trained model in `result_files/saved_models/gb1None_DD_MM_YYYY_HHMMSS/` as well as the used parameters in `result_files/log_file.csv` and the training and test statistics in `result_files/results.csv`.
 
 ### Training without using a pre-trained model
-`python3 d4_cmd_driver.py --query_name gb1 --alignment_file example/gb1_1000_experimental.clustal --tsv_filepath example/gb1None.tsv --pdb_filepath example/gb1.pdb --number_mutations num_mutations --variants variant --score score --wt_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --first_ind 0 --training_epochs 100 --split0 0.7 --split1 0.15 --split2 0.05 --save_model`
+```sh
+python3 d4_cmd_driver.py --query_name gb1 --alignment_file example/gb1_1000_experimental.clustal --tsv_filepath example/gb1None.tsv --pdb_filepath example/gb1.pdb --number_mutations num_mutations --variants variant --score score --wt_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --first_ind 0 --training_epochs 100 --split0 0.7 --split1 0.15 --split2 0.05 --save_model
+```
 
 ### Training by using a pre-trained model
-`python3 d4_cmd_driver.py --query_name gb1 --alignment_file example/gb1_1000_experimental.clustal --tsv_filepath example/gb1None.tsv --pdb_filepath example/gb1.pdb --number_mutations num_mutations --variants variant --score score --wt_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --first_ind 0 --training_epochs 100 --split0 0.7 --split1 0.15 --split2 0.05 --save_model --transfer_conv_weights result_files/saved_models/gb1None_DD_MM_YYYY_HHMMSS/ --train_conv_layers`
+*CAUTION check the correct time stamp for the pre-trained model*
+```sh
+python3 d4_cmd_driver.py --query_name gb1 --alignment_file example/gb1_1000_experimental.clustal --tsv_filepath example/gb1None.tsv --pdb_filepath example/gb1.pdb --number_mutations num_mutations --variants variant --score score --wt_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --first_ind 0 --training_epochs 100 --split0 0.7 --split1 0.15 --split2 0.05 --save_model --transfer_conv_weights result_files/saved_models/gb1None_DD_MM_YYYY_HHMMSS/ --train_conv_layers
+```
 
 ## Making predictions
+*CAUTION check the correct time stamp for the trained model*
+
 To make predictions for the variants `G1W` as well as `K42G,I24P` and get the predicted fitness score.
-`d4_predict.py --model_filepath result_files/saved_models/gb1None_10_01_2024_110517_end/ --protein_pdb example/gb1.pdb --alignment_file example/gb1_1000_experimental.clustal --query_name gb1 --variant_s G1W_K42G,I24P --wt_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --first_ind 0`
-(`| awk -F '_' '{print $1} $2 > 0' | grep _` can be added if only positive scores shoud be shown [assumes a UNIX machine])
 
+```sh
+d4_predict.py --model_filepath result_files/saved_models/gb1None_10_01_2024_110517_end/ --protein_pdb example/gb1.pdb --alignment_file example/gb1_1000_experimental.clustal --query_name gb1 --variant_s G1W_K42G,I24P --wt_seq MQYKLILNGKTLKGETTTEAVDAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE --first_ind 0
+```
+
+If only positive scores should be shown, one can pipe the output through the following commands:
+
+```bash
+| awk -F '_' '{print $1} $2 > 0' | grep _
+```
+ can be added if only positive scores shoud be shown (assumes a UNIX machine)