File size: 3,466 Bytes
945e89a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ddcd55c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
---
dataset_info:
  features:
  - name: input_ids
    sequence: int32
  - name: coords
    sequence:
      sequence: float64
  - name: labels
    dtype: int64
  - name: Dgscore
    dtype: float64
  - name: token_type_ids
    sequence: int8
  splits:
  - name: train
    num_bytes: 133082101
    num_examples: 304
  - name: val
    num_bytes: 42691055
    num_examples: 110
  - name: test
    num_bytes: 55786244
    num_examples: 104
  download_size: 58746710
  dataset_size: 231559400
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
  - split: test
    path: data/test-*
---

# LEP: Ligand Efficacy Prediction

## Overview

To design drugs with desired effects, the task is to predict whether a given drug molecules would be activators or inactivators of a given protein. We directly incorporate the structure of both the active and inactive proteins into the learning task in addition to the drug molecule.

The dataset consists of 29 proteins from several families with both "active" and "inactive" state structures. For example, an open or closed ion channel or a GPCR bound to a G protein or apo. We then modeled in 527 small molecules with known activating or inactivating function and likely binding site using the program Glide (selecting the best predicted binding pose of the small molecule). 

We then input this pair structures for a particular molecule, and attempt to classify the molecule as activator or inactivator (A or I) of that protein's function. This can also be viewed as approximately predicting which protein conformation the ligand will bind more strongly to. 


## Datasets
- splits:
   - split-by-protein: all complexes with the same protein are in the same split dataset


## Format

Each entry in the dataset contains the following keys:

    ['input_ids'] The set of atomic numbers for both the active and inactive structures concatenated together.
    ['coords'] The 3D coordinates for both the active and inactive structures concatenated together.
    ['labels'] 1 for activator, and 0 for inactivator.
    ['Dgscore'] (float) Difference of the Glide scores (see below)
    ['token_type_ids'] A mask that corresponds to which inputs_ids/coords belong to the active or inactive structure (0 for active , 1 for inactive)


As a baseline measure, the difference in the best Glide scores of the ligand to each structure is included (Dgscore). This can be used as a baseline predictor of ligand activity in this context.

## Citation Information

```
@article{townshend2020atom3d,
  title={Atom3d: Tasks on molecules in three dimensions},
  author={Townshend, Raphael JL and V{\"o}gele, Martin and Suriana, Patricia and Derry, Alexander and Powers, Alexander and Laloudakis, Yianni and Balachandar, Sidhika and Jing, Bowen and Anderson, Brandon and Eismann, Stephan and others},
  journal={arXiv preprint arXiv:2012.04035},
  year={2020}
}
```

```
@article{friesner2004glide,
  title={Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy},
  author={Friesner, Richard A and Banks, Jay L and Murphy, Robert B and Halgren, Thomas A and Klicic, Jasna J and Mainz, Daniel T and Repasky, Matthew P and Knoll, Eric H and Shelley, Mee and Perry, Jason K and others},
  journal={Journal of medicinal chemistry},
  volume={47},
  number={7},
  pages={1739--1749},
  year={2004},
  publisher={ACS Publications}
}
```