Datasets:
monsoon-nlp
commited on
Commit
•
233e20a
1
Parent(s):
e52e9c0
Update README.md
Browse files
README.md
CHANGED
@@ -5,4 +5,22 @@ task_categories:
|
|
5 |
tags:
|
6 |
- biology
|
7 |
- protein
|
8 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
tags:
|
6 |
- biology
|
7 |
- protein
|
8 |
+
---
|
9 |
+
|
10 |
+
# Protein Pairs and Similarity
|
11 |
+
|
12 |
+
Selected protein similarities within training, test, and validation sets.
|
13 |
+
Each protein gets two similarities selected at random and (if possible) proteins within the top and bottom quintiles for similarity.
|
14 |
+
|
15 |
+
The protein is represented by its UniProt ID and its amino acid sequence (using IUPAC-IUB codes where each amino acid maps to a letter of the alphabet, see: https://en.wikipedia.org/wiki/FASTA_format ).
|
16 |
+
|
17 |
+
The distance column is cosine distance (identical = 0) on UniProt / SwissProt 1,024-dimension embeddings, downloaded in March 2024.
|
18 |
+
For models and training scripts optimizing for similarity, you should use `(1 - distance)`.
|
19 |
+
|
20 |
+
Train / test / validation splits are from [khairi/uniprot-swissprot](https://huggingface.co/datasets/khairi/uniprot-swissprot/tree/39d2f9010ceb2599a12397b79babc902f8c440e6)
|
21 |
+
|
22 |
+
Code notebook: https://colab.research.google.com/drive/1rhoF0pAauHbiaNHdZkOI5O7A3mMcvS6G?usp=sharing
|
23 |
+
|
24 |
+
Source of proteins: UniProt-SwissProt
|
25 |
+
|
26 |
+
Download 1,024-dim embeddings from UniProt: https://www.uniprot.org/help/downloads#embeddings
|