Release notes

Recently published apps

We have just published DRAGMAP, the open source DRAGEN mapper/aligner that can be used to align single or paired-end reads (FASTQ) or an input BAM file. The app is available in the Public Apps gallery.

Recently published apps

We have just updated the content of our public app galleries with new GATK releases:

• GATK Pre-Processing For Variant Discovery 4.2.0.0 workflow is used to prepare data for variant calling analysis. The workflow consists of two major segments: alignment to reference genome and data cleanup operations that correct technical biases. Resulting BAM files are ready for variant calling analysis and can be further processed by other BROAD best practice pipelines, like Generic Germline Short Variant Per-Sample Calling workflow, Somatic CNVs workflow, and Somatic SNVs + INDELs workflow.
• GATK Generic Germline Short Variant Per-Sample Calling 4.2.0.0 workflow that calls germline variants in a WGS sample with GATK HaplotypeCaller, starting from an analysis-ready BAM file.

And six GATK 4.2.0.0 tools:

• GATK GatherBQSRReports tool that gathers scattered BQSR recalibration reports into a single file.
• GATK BaseRecalibrator tool that generates a recalibration table based on various covariates for input mapped read data.
• GATK ApplyBQSR tool that recalibrates the base quality scores of an input BAM or CRAM file containing reads.
• GATK HaplotypeCaller tool for calling germline SNPs and indels from input BAM file(s) via local re-assembly of haplotypes.
• GATK VariantFiltration tool used for filtering variants in a VCF file based on INFO and/or FORMAT annotations.
• GATK MergeVcfs, used for combining multiple variant files.

Recently published apps

We’ve just published four tools from the OncoGEMINI 1.0.0 toolkit:

• OncoGEMINI Bottleneck that identifies somatic variants with increasing allele frequency in longitudinal data.

• OncoGEMINI Loh, a command tool that performs loss of heterozygosity analysis.

• OncoGEMINI Truncal that recovers variants that appear in all tumor samples, but are absent in the normal sample.

• OncoGEMINI Unique tool for identifying somatic variants unique to a subset of samples.

Billing information just got more informative and organized

The following improvements have been made on the Billing page available to Enterprise and Division administrators:

• The Billing page has been redesigned and now consists of three sections: Billing information, Instance limits and Payment information.
• The start date from which costs are calculated is now displayed for the current billing period.
• Additional charges and credits information is now renamed to Charges and Refunds and grouped in the Additional subsection.
• Total Platform charges now sum up costs for Analysis, Storage and Additional charges.

Recently published apps

SBG Image Processing Toolkit 

SBG Image Processing Toolkit consists of apps that enable various stages of machine learning image processing. Seamless integration between the tools of this toolkit provides an easy and logical analysis flow, while enabling support of various data types, preprocessing steps and utilizing computation capabilities of the Seven Bridges Platform.

1. SBG Deep Learning Image Classification Exploratory Workflow is an image classifier pipeline that relies on the transfer learning approach. This allows the use of pre-trained models as the starting point for building a model adjusted to given image datasets. Furthermore, the pipeline allows training of the model for a variety of hyperparameter combinations in parallel by utilizing multiple GPU instances, while detailed metrics and visualizations help determine the best configuration that can later be used to make predictions on new data instances. 
2. SBG Deep Learning Prediction is an image classifier tool that classifies unlabeled images based on labeled data. It is intended as a final step after the SBG Deep Learning Image Classification Exploratory Workflow. Testing different configurations in parallel with the exploratory workflow and finding the best model configuration for the given dataset, then using SBG Deep Learning Prediction with that configuration and all available labeled images as the training data provides the optimal training conditions which lead to the best classification results.
3. SBG Histology Whole Slide Image Preprocessing takes SVS histopathology images, removes various artifacts, and outputs the desired number of best quality tiles in PNG format that consist of at least 90% tissue.
4. SBG X-Ray Image Preprocessing Workflow performs the selected X-ray image enhancement algorithm: unsharp masking (UM), high-frequency emphasis filtering (HEF) or contrast limited adaptive histogram equalization (CLAHE). 
5. SBG Stain Normalization involves casting an array of images in the stain colors of a target image. Stain normalization is used as a histopathology image preprocessing step to reduce the color and intensity variations present in stained images obtained from different laboratories.
6. SBG Medical Image Convert performs medical image format conversion. If the input data are medical images in a non-standard format (e.g. SVS, TIFF, DCM or DICOM), SBG Medical Image Convert converts them to PNG format.
   
7. SBG Split Folders organizes an image directory into the train and test subdirectory structure. These directories are necessary inputs for SBG Deep Learning Image Classification Exploratory Workflow and SBG Deep Learning Prediction.

HistoQC

HistoQC is an open-source quality control tool for digital pathology slides. It performs fast quality control to not only identify and delineate artefacts but also discover cohort-level outliers (e.g., slides stained darker or lighter than others in the cohort). It outputs an interactive user interface for easy viewing and understanding of the results.

Minimac4

Minimac4 is a genetic imputation algorithm that can be used to impute genotypes in a genomic region starting from a reference panel in M3VCF format and pre-phased target GWAS haplotypes.

BOLT-LMM

BOLT-LMM is a tool that tests the association between genotypes and phenotypes using a linear mixed model.

Recently published apps

GSEAPreranked Workflow performs Gene Set Enrichment Analysis (GSEA). It is generated with an assumption that a differential expression analysis has been done before using the DESeq2 tool which is publicly available on the Seven Bridges Platform. The GSEAPreranked Workflow consists of two tools, GSEA Input Prepare and GSEAPreranked. The GSEAPreranked tool represents a wrapper around the command-line tool that was developed by the BROAD Institute. The GSEA Input Prepare tool is based on the Python script developed by the Seven Bridges team to prepare the required input file formats for the GSEAPreranked tool.

Amazon EC2 GPU G4dn instances available on the Platform

With this update you can now use the newest Amazon EC2 GPU G4dn instances, in task executions and Data Cruncher analyses, as the industry’s most cost-effective and versatile GPU instances for deploying machine learning models. 

G4dn instances feature NVIDIA T4 GPUs and custom Intel Cascade Lake CPUs, and are optimized for machine learning inference and small scale training. 

NVIDIA drivers come preinstalled and optimized according to the Amazon best practice for the specific instance family and are accessible from the Docker container.

The following instances have been added:

• g4dn.xlarge
• g4dn.2xlarge
• g4dn.4xlarge
• g4dn.8xlarge
• g4dn.16xlarge
• g4dn.12xlarge

Learn more about supported GPU instance types.

Recently published apps

ENCODE ChIP-Seq Pipeline 2 analysis studies chromatin modifications and binding patterns of transcription factors and other proteins. It combines chromatin immunoprecipitation (ChIP) assays with standard NGS sequencing. The workflow is based on ChIP-Seq 2 pipeline, developed by the ENCODE Consortium.

ENCODE ATAC-seq Pipeline performs quality control and signal processing, producing alignments and measures of enrichment. The Assay for Transposase-Accessible Chromatin followed by sequencing (ATAC-seq) experiment provides genome-wide profiles of chromatin accessibility. Briefly, the ATAC-seq method works as follows: loaded transposase inserts sequencing primers into open chromatin sites across the genome, and reads are then sequenced. The ends of the reads mark open chromatin sites. The workflow is based on the ENCODE ATAC-seq pipeline, developed by the ENCODE Consortium.

SB CLI now accepts automation start parameters as YML or JSON file

Starting from SB CLI version 0.18.0, arguments necessary for initiating a new automation run via the automations CLI can come from inside a user-provided YML or JSON file. 

This allows you to maintain automation parameters inside well organized and annotated file templates, instead of providing all arguments (in particular automation inputs, settings, and secret settings) in the form of a long, opaque command line string that is difficult to read and write.

When starting a new automation, you would first need to modify the local YML/JSON file and then provide the file path as an argument to the sb automations start command. Read more on the Manage via CLI RHEO documentation page.

GDC Datasets version update

As of May 27, GDC datasets available through the Data Browser and the API correspond to GDC Data Release 29.0.