cleanup

.gitignore

@@ -1,13 +1,8 @@
 **/*.pyc
 **/__pycache__
 .ipynb_checkpoints
-old
 
 data/**
 !data/README.md
 
-pretrained/**
-!pretrained/README.md
-
 .DS_Store
-grid_0.right_answer.json

README.md

@@ -1,24 +1,22 @@
 
-# Lead Optimization
+# DeepFrag
 
 This repository contains code for machine learning based lead optimization.
 
 # Overview
 
 - `config`: fixed configuration information (eg. TRAIN/VAL/TEST partitions)
-- `configurations`: benchmark model configurations
+- `configurations`: benchmark model configurations (see [`configurations/README.md`](configurations/README.md))
 - `data`: training/inference data (see [`data/README.md`](data/README.md))
 - `leadopt`: main module code
     - `models`: pytorch architecture definitions
     - `data_util.py`: utility code for reading packed fragment/fingerprint data files
     - `grid_util.py`: GPU-accelerated grid generation code
-    - (outdated) `infer.py`: code for inference with a trained model
     - `metrics.py`: pytorch implementations of several metrics
     - `model_conf.py`: contains code to configure and train models
     - `util.py`: utility code for rdkit/openbabel processing
-- (outdated) `scripts`: data processing scripts (see [`scripts/README.md`](scripts/README.md))
+- `scripts`: data processing scripts (see [`scripts/README.md`](scripts/README.md))
 - `train.py`: CLI interface to launch training runs
-- (outdated) `leadopt.py`: CLI interface to run inference on new samples
 
 # Dependencies
 

configurations/README.md

@@ -1,6 +1,11 @@
 This folder contains benchmark model configurations referenced in the paper.
 
-You can retrain from these configurations using the `train.py` script:
+Overview:
+- `layer_type_sweep/*`: experimenting with different parent/receptor typing schemes
+- `voxelation_sweep/*`: experimenting with different voxelation types and atomic influence radii
+- `final.json`: final production model
+
+You can train new models using these configurations with the `train.py` script:
 
 ```sh
 python train.py \

configurations/final.json

@@ -0,0 +1,35 @@
+{
+    "version": "voxelnet",
+    "no_partitions": false,
+    "fragments": "./data/moad.h5",
+    "fingerprints": "./data/rdk10_moad.h5",
+    "learning_rate": 0.001,
+    "num_epochs": 50,
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24,
+    "grid_res": 0.75,
+    "fdist_min": null,
+    "fdist_max": 4,
+    "fmass_min": null,
+    "fmass_max": 150,
+    "ignore_receptor": false,
+    "ignore_parent": false,
+    "output_size": 2048,
+    "pad": false,
+    "blocks": [
+        64,
+        64
+    ],
+    "fc": [
+        512
+    ],
+    "use_all_labels": true,
+    "dist_fn": "cos",
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 0,
+    "rec_typer": "simple",
+    "acc_type": 0,
+    "lig_typer": "simple"
+}

configurations/layer_type_sweep/lig_simple.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "simple",
+    "acc_type": 0,
+    "lig_typer": "simple"
+}

configurations/layer_type_sweep/lig_simple_h.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "simple",
+    "acc_type": 0,
+    "lig_typer": "simple_h"
+}

configurations/layer_type_sweep/lig_single.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "simple",
+    "acc_type": 0,
+    "lig_typer": "single"
+}

configurations/layer_type_sweep/lig_single_h.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "simple",
+    "acc_type": 0,
+    "lig_typer": "single_h"
+}

configurations/layer_type_sweep/rec_meta.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "meta",
+    "acc_type": 0,
+    "lig_typer": "simple"
+}

configurations/layer_type_sweep/rec_meta_mix.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "meta_mix",
+    "acc_type": 0,
+    "lig_typer": "simple"
+}

configurations/layer_type_sweep/rec_simple_h.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "simple_h",
+    "acc_type": 0,
+    "lig_typer": "simple"
+}

configurations/layer_type_sweep/rec_single.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "single",
+    "acc_type": 0,
+    "lig_typer": "simple"
+}

configurations/layer_type_sweep/rec_single_h.json

@@ -0,0 +1,30 @@
+{
+    "version": "voxelnet", 
+    "no_partitions": false, 
+    "fragments": "./data/moad.h5", 
+    "fingerprints": "./data/rdk10_moad.h5", 
+    "learning_rate": 0.001, 
+    "num_epochs": 15, 
+    "test_steps": 400,
+    "batch_size": 16,
+    "grid_width": 24, 
+    "grid_res": 0.75, 
+    "fdist_min": null, 
+    "fdist_max": 4, 
+    "fmass_min": null, 
+    "fmass_max": 150, 
+    "ignore_receptor": false, 
+    "ignore_parent": false, 
+    "output_size": 2048, 
+    "pad": false, 
+    "blocks": [64, 64], 
+    "fc": [512], 
+    "use_all_labels": true, 
+    "dist_fn": "cos", 
+    "loss": "direct",
+    "point_radius": 1,
+    "point_type": 3,
+    "rec_typer": "single_h",
+    "acc_type": 0,
+    "lig_typer": "simple"
+}