Compute Environments

Every Drylab sandbox starts from a prebuilt software environment, so you don't have to install the usual bioinformatics stack yourself. Drylab installs anything extra it needs during the analysis.

Cloud sandboxes

Cloud instances run Drylab's comprehensive research image. You don't need to choose an image; Drylab picks the right one for the instance you select.

The image covers a broad range of work:

• Multi-omics: bulk and single-cell RNA-seq, genomics and variant analysis, epigenomics (ChIP-seq, ATAC-seq, methylation), proteomics and metabolomics, pathway analysis.
• Single-cell and spatial: QC, integration, annotation, trajectories, RNA velocity, spatial transcriptomics.
• Imaging: microscopy segmentation and tracking, histology and whole-slide images.
• Flow and mass cytometry: preprocessing, gating, clustering.
• Chemistry and drug discovery: molecular descriptors, QSAR, docking preparation.
• Machine learning and statistics: classical ML, gradient boosting, Bayesian modeling, model interpretation.
• Workflows: Snakemake and nf-core/Nextflow for reproducible pipelines.

Included tooling (examples)

Python

• Core: NumPy, pandas, SciPy, scikit-learn, Matplotlib, Seaborn, Plotly
• Single-cell and spatial: Scanpy, AnnData, scvi-tools, CellRank, scVelo, Squidpy, cell2location
• Imaging: Cellpose, StarDist, napari
• Genomics: pysam, Biopython, pybedtools, pyranges, cyvcf2
• ML: XGBoost, LightGBM, PyMC, SHAP
• Workflows: Snakemake, nf-core tools

R

• Bioconductor: DESeq2, edgeR, limma, SingleCellExperiment, scran, scater
• Single-cell: Seurat and many companion packages
• Flow cytometry: flowCore, FlowSOM
• Reporting: tidyverse, ggplot2, rmarkdown

Command line

• Alignment, trimming and QC: STAR, Bowtie2, cutadapt, fastp, MultiQC
• Variant and interval tools: samtools, bcftools, bedtools
• Peak calling: MACS3
• Workflows: Snakemake, Nextflow, nf-core tools

Some aligners, such as HISAT2, BWA-MEM2 and minimap2, aren't in the sandbox itself. Drylab runs them as accelerated tools on dedicated hardware, or installs them when you ask.

Heavy steps

Your sandbox doesn't have to do all the work. Heavy steps, such as structure prediction, read alignment or clustering a large single-cell atlas, can run on Drylab's GPU and HPC hardware while the rest of the analysis stays in your sandbox. See Heavy steps run on GPU automatically.

Local and Desktop runtimes

When you run Drylab on your own machine, you choose the image when you set up the runtime:

See Local Runtime and Install the Desktop App.