HumanBase Beluga

Beluga analysis
Beluga is a deep learning framework for predicting genomic variant effects on regulatory features using a convolutional neural network trained on chromatin profiling data.

Submit a Beluga analysis

Model:

Beluga (2019 DNA model)

Type:

VCF

Genome assembly:

GRCh37/hg19

Show format descriptions

VCF (Variant effect prediction)

Upload a VCF file for variant effect prediction. VCF format:

Each line must have at least 5 tab separated fields.
If there is a header column, it must begin with a hash `#` and the first 5 tab separated entries must be the following: `#CHROM`, `POS`, `ID`, `REF`, `ALT`
The first column: Chromosome name
The second column: Position. A positive integer within chromosome length
The third column: Variant name. This field can be left empty.
The fourth column: Reference allele. One or more upper case codes from ATCG. Maximum 100 characters.
The fifth column: Alternative allele. Contain one or more upper case codes from ATCG. Maximum 100 characters.
Required for Seqweaver - The sixth column: Strand. Either '+' or '-' indicating strand.

A VCF file can be opened in any text editor. However, for most users a VCF file is best viewed using a spreadsheet program such as Microsoft Excel or Google Sheets.

Example VCF file
Example VCF file with strand

BED (Chromatin feature prediction)

Upload a BED file for chromatin feature prediction. BED format:

Each line has at least 3 tab separated fields.
The first column: Chromosome name
The second column: Start Position. A non-negative integer within chromosome length
The third column: End Position. A non-negative integer within chromosome length
Center Position Analysis: We analyze the center position of the interval. For even-length intervals, we use the left-center position (floor division). Examples:
- chr1 100 100 (equal coordinates) → center at position 100
- chr1 100 101 (single base) → center at position 100
- chr1 100 105 (5 bases) → center at position 102
- chr1 100 106 (6 bases) → center at position 103 (left-center)

Example BED file

FASTA (chromatin feature prediction / in-silico mutagenesis)

Upload a FASTA file for chromatin feature prediction or in-silico mutagenesis. FASTA format:

Each section starts with ">".
Sequence length depends on the selected model: Sei (4096bp), Beluga (2000bp), or Seqweaver (1000bp)
Allowed nucleic acid codes: ATCGatcg (lowercase is treated the same of uppercase).

Example FASTA file

Upload Options:

File upload

Paste contents

Example vcf submission

Select an upload file:

Job Title (optional):

Enter email for results (optional):

Beluga is a deep learning framework for predicting genomic variant effects on regulatory features using a convolutional neural network trained on chromatin profiling data.

Visitfor additional help and details.

Flatiron Institute - Simons Foundation | How to Cite

Beluga analysisBeluga is a deep learning framework for predicting genomic variant effects on regulatory features using a convolutional neural network trained on chromatin profiling data.

Submit a Beluga analysis

VCF (Variant effect prediction)

BED (Chromatin feature prediction)

FASTA (chromatin feature prediction / in-silico mutagenesis)

Beluga analysis
Beluga is a deep learning framework for predicting genomic variant effects on regulatory features using a convolutional neural network trained on chromatin profiling data.