File size: 1,761 Bytes
c36421a
f00ed2f
 
 
 
 
c36421a
f00ed2f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: cc-by-nc-4.0
tags:
- genomics
- plant-genomics
- pytorch
---

# DCGP model checkpoints

DCGP (Domestication Convergence Gene Prediction) predicts candidate genes under
convergent selection during plant domestication. This repository contains 44
trained PyTorch classifier checkpoints. The classifiers use sequence embeddings
from Evo 2. See the [DCGP code repository](https://github.com/AIBreeding/DCGP)
for preprocessing, inference, and environment setup.

## Download and run

Each checkpoint is stored as `model/<model-id>/<model-id>.pt`. Select the
model ID for your analysis. For example, from the root of the DCGP code repository:

```bash
hf download AIBreeding/DCGP model/Brassica_napusA/Brassica_napusA.pt --local-dir .
python embedding.py --fasta sequence/sample.fa --step 250 --window 1024 \
  --layer blocks.24 --out sequence/sa.pt
python predict.py --data sequence/sa.pt \
  --model model/Brassica_napusA/Brassica_napusA.pt --out results/pred
```

The embedding step uses `evo2_7b` and requires CUDA and SeqKit. The predictor
accepts precomputed embeddings and writes scores and predicted labels to
`results/pred/predict_result.csv`. For reproducible runs, pass the full Hub
commit SHA using `hf download --revision`. File hashes are listed in
`SHA256SUMS`.

## Intended use

These checkpoints are intended for noncommercial research and reproduction of
the DCGP inference workflow. Predicted scores can help prioritize candidates
for follow-up analysis and experimental validation.

## License

The DCGP checkpoints are available under
[Creative Commons Attribution-NonCommercial 4.0 International](https://creativecommons.org/licenses/by-nc/4.0/legalcode.en).
The source code and upstream Evo 2 weights have their own license terms.