DLEB

Data Download

Users can download various example biological datasets for training and testing their deep learning models. DLEB also provides example Python scripts of deep learning models for each data type.

Example sequence data

expand_more The seqeunces of transcript factor binding sites (TFBSs) in human genome

TFBS sequence data 439 Mb Example model 8.4 Kb

Go to edit example model arrow_forward_ios

Description

The data was used for the study Zhou, Jian, and Olga G. Troyanskaya. "Predicting effects of noncoding variants with deep learning–based sequence model." Nature methods 12.10 (2015): 931-934.

The human genome (hg19) is split into 1000-bp bins, and the labels were computed for all 919 chromatin features for each bin.

Dataset

The forward and complementary sequences of 1000bp bins with at least one transcription factor (TF) binding event (FASTA format)

Label data

Chromatin features for each bin

Example mapping data

expand_more The mapping data of mouse sequencing reads to mouse genome assembly

SNP genotype data 6.9 Gb INDEL genotype data 2.4 Gb Count data 6.9Gb Example model (Count data) 8.2Kb

Go to edit example model arrow_forward_ios

Description

The data was used for the study Poplin, Ryan, et al. "A universal SNP and small-indel variant caller using deep neural networks." Nature biotechnology 36.10 (2018): 983-987.

The original BAM file for alignment of mouse (129S1_SvImJ) sequencing reads to mouse reference genome (mm10) and VCF file for SNP and INDEL variants of 129S1_SvImJ were obtained from the above paper. The mapping information corresponding to bins used as labels was only extracted and used to build example datasets.

Dataset

The mapping results of sequencing reads in 201bp bins (BAM format)

Label data 1

The genotype of variant located in the center of each bin with the position of the bin (BED format)

Label data 2

The number of variants contained in each bin (BED format)

Example profile (signal) data

expand_more The DNase-seq and histone modification profiles of human

Chromatin annotation data 2.8 Gb Example model 10 Kb

Go to edit example model arrow_forward_ios

Description

The data was used for the study Zhou, Jian, and Olga G. Troyanskaya. "Ernst, Jason, and Manolis Kellis. "ChromHMM: automating chromatin-state discovery and characterization." Nature methods 9.3 (2012): 215-216.

The DNase-seq profile (ENCFF264NMW), histone modification profiles (ENCFF180LKW, ENCFF682WPF, ENCFF828CQV, ENCFF818GNV, ENCFF465KNK) and chromatin annotation (ENCFF001TDH) of GM12878 cell line were obtained from ENCODE project. The human genome (hg19) was split into 200-bp bins, and the bins whose all bases were covered by the same chromatin annotation were only used for label datasets. The profile information corresponding to bins used as labels were extracted and used to build example datasets.

Dataset

The DNase-seq and 5 different Histone modification profiles (H3K27ac, H3K4me1, H3K4me2, H3K4me3, H3K9ac) of 200bp bins (bigWig format)

Label data

Chromatin features for each bin (BED format)

Example image data

expand_more The in situ hybridization (ISH) images of Drosophila embryos in different developmental stages

ISH image data 307 Mb Example model 8.8 Kb

Go to edit example model arrow_forward_ios

Description

The data was used for the study Cai, Lei, et al. "Deep Low-Shot Learning for Biological Image Classification and Visualization From Limited Training Samples." IEEE Transactions on Neural Networks and Learning Systems (2021).

The ISH images in the lateral direction of Drosophila embryos on 6 different developmental stages were obtained from the Berkeley Drosophila Genome Project (BDGP). The obtained images were cropped into 128 x 320 size.

Dataset

The 128 x 320 sized ISH images of diverse genes in the lateral direction

Label data

The developmental stage of each image