File size: 1,626 Bytes
542aafd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: cc-by-4.0
---

# Nanobody (VHH) Affinity Prediction Dataset

## Dataset Overview

This dataset helps predict the binding affinity between nanobodies (VHH, single-domain antibodies from camelids) and their target antigens. Affinity is a key parameter that measures how strongly an antibody binds to its antigen, usually expressed as dissociation constant (KD) or binding free energy.

High affinity is a critical property for therapeutic antibodies, so accurately predicting nanobody affinity is important for antibody engineering and screening.

## Data Collection

The dataset is based on experimentally measured nanobody-antigen binding affinities. Data is collected from published literature and split based on score (stratified split)

## Dataset Structure

The dataset is split into training, validation, and test sets.

### File Format

CSV files contain these columns:
- `seq`: Nanobody amino acid sequence
- `score`: Affinity value (typically -log10(KD) where KD is in M), higher values indicate stronger binding affinity

## Uses and Limitations

### Uses
- Develop models to predict nanobody affinity
- Help select and optimize nanobodies
- Reduce experimental work and speed up drug development

### Limitations
- Differences in affinity measurement methods may cause data variability
- The same antibody-antigen pair may have different affinity values under different conditions
- The dataset may not cover all possible nanobody-antigen combinations

## Evaluation Metrics

Model performance is evaluated using:
- Spearman correlation
-- Root Mean Squared Error (RMSE)
- Mean Absolute Error (MAE)