kuleshov-group/cross-species-single-nucleotide-annotation

Dataset Overview

The dataset consists of five tasks for cross-species modeling plant genomes at single-nucleotide resolution in plants. These tasks are:

1. Translation Initiation Site (TIS) Prediction
2. Translation Termination Site (TTS) Prediction
3. Splice Donor Site Prediction
4. Splice Acceptor Site Prediction
5. Evolutionary Conservation Prediction

Tasks 1-4: Site Predictions

• Training Dataset: Generated from Arabidopsis chromosomes 1-4
• Validation Dataset: Generated from Arabidopsis chromosome 5
• Testing Datasets: Compiled from rice, sorghum, and maize

Task 5: Evolutionary Conservation Prediction

• Training Dataset: Generated from sorghum chromosomes 1-9
• Validation Dataset: Generated from sorghum chromosome 10
• Testing Datasets: Compiled in maize

This datasets facilitate robust cross-species nucleotide annotation.

Dataset sizes

TIS, TTS, Donor, Acceptor

	TIS		TTS		Donor		Acceptor
	# of positives	# of negatives	# of positives	# of negatives	# of positives	# of negatives	# of positives	# of negatives
Training on Arabidopsis chromosome 1-4	24,711	173,880	25,112	220,452	96,752	483,268	97,224	536,179
Validation on Arabidopsis chromosome 5	7,311	50,514	7,461	64,365	29,377	140,536	29,567	155,397
Rice test	2,974	1,400,115	2,974	3,718,029	21,963	3,764,549	21,963	4,151,774
Sorghum test	3,214	3,937,719	3,214	10,445,530	24,801	10,821,941	24,801	12,640,573
Maize test	3,098	11,265,574	3,098	29,535,973	24,399	34,516,038	24,399	40,025,899

Evolutionary conservation

	# of positives	# of negative
Train	429,043	429,043
Validation	19,030	19,030
Test	947,769	976,230

How to use

from datasets import load_dataset
import pandas as pd
repo_id = 'kuleshov-group/cross-species-single-nucleotide-annotation'
tis = load_dataset(repo_id, data_files={'train': 'TIS/train.tsv', 'valid': 'TIS/valid.tsv', 'test_rice':'TIS/test_rice.tsv', 'test_sorghum':'TIS/test_sorghum.tsv', 'test_maize':'TIS/test_maize.tsv'})
tis_train = tis['train']
# convert to pandas dataframe
tis_train_df = tis_train.to_pandas()