Drylab

Chat

Mention Files and Resources

Type @ in the chatbox to point Drylab at your files and at built-in tools, workflows, pipelines, databases and atlases.

Type @ in the chatbox to open the resource picker. Mentioning a resource tells Drylab exactly which file, tool or dataset to use, which gives faster and more accurate results than describing it in words.

Why mention instead of describe

Compare "load my counts file" with "load @counts.csv":

  • No guessing. Drylab uses the exact file, folder or tool you picked, even if several have similar names.
  • The right tool, first time. Mentioning Boltz2 or an nf-core pipeline makes Drylab use that tool instead of choosing one itself.
  • Faster runs. Drylab doesn't spend time searching your project or the catalogue, so it starts working sooner.
  • Easier to reread. Mentions appear as colored pills in your message, so you can see at a glance what each analysis used.

The @ picker: files in this chat and recent files in the project, with the category tabs on top

When you type just @, the picker starts with the files you're most likely to need: In this chat (files the chat used or made) and Recent in project. Type to search everything else.

What you can mention with @: files, category shortcuts, keyboard, built-in resources and Your Data

Categories

The picker has one tab per category. Once you type, each tab shows how many items match. Press Tab to move to the next tab and Shift + Tab to go back.

TabWhat it listsMention it when you want to...
AllEverything below in one listSearch across everything
FilesFiles in the current projectAnalyze a specific file
FoldersFolders in the current projectWork on a whole folder, such as a run of FASTQ files
ToolsAccelerated tools such as Boltz2, Chai-1 or STARUse a particular tool, often on a GPU
WorkflowsGuided, step-by-step analyses (see Structured Workflows)Run a validated analysis that asks you at the key decisions
Pipelinesnf-core pipelines for raw sequencing dataProcess raw reads with a standard pipeline
DatabasesBiological databases such as UniProt, ClinVar or CellxGeneLook up or pull reference data
AtlasCurated public datasets that Drylab reads in place, with no download (see Drylab Atlas)Compare your data with a published dataset
Your DataCloud storage buckets you connected in Storage > Data ConnectorAnalyze data that stays in your own bucket

What's built in

Besides your own files, the picker gives you Drylab's built-in library. You don't install or configure any of it.

Databases

130+ biological databases you can query in plain language, without API keys or scripts: gene and protein information, structures, variants, pathways, drugs and expression.

CategoryExamples
Proteins and structuresUniProt, PDB, AlphaFold, CATH
Genes and variantsEnsembl, gnomAD, ClinVar, 1000 Genomes, CIViC
Pathways and interactionsKEGG, Reactome, BioGRID
Drugs and chemistryChEMBL, PubChem, BindingDB
CancercBioPortal
Expression and single-cellGTEx, GEO, ArrayExpress, CellxGene

Example: "Query @ClinVar for pathogenic BRCA1 variants and summarize them by exon."

Accelerated tools

About 250 tools run heavy jobs (structure prediction, docking, protein design, alignment, assembly and more) on dedicated GPU and HPC hardware and bring the results back to your chat. See Accelerated Tools for the full list, costs and best practices.

Pipelines

About 80 production nf-core pipelines take you from raw FASTQ files to analysis-ready outputs with standardized, reproducible processing. The jobs run on Drylab's HPC hardware; follow them in HPC Tasks.

CategoryExamples
Bulk RNA-seqnf-core Rnaseq, nf-core Rnavar
Single-cellnf-core Scrnaseq
Variant callingnf-core Sarek
Amplicon and 16Snf-core Ampliseq
ChIP-seq and ATAC-seqnf-core Chipseq, nf-core Atacseq
Metagenomicsnf-core Mag, nf-core Taxprofiler

Workflows, Atlas and skills

  • Workflows combine several steps into a guided analysis that stops for your decisions. See Structured Workflows.
  • Atlas datasets are curated public studies that Drylab reads in place. See Drylab Atlas.
  • Drylab also follows built-in best-practice protocols (skills) on its own when they match your task. They aren't in the picker. To make it follow your lab's own rules, mention a protocol file.

Find what you need

Keep typing after @ to filter. The search matches names, descriptions and keywords, so related items can appear too. On All, the best result shows first as Top match, followed by a few results per category; click See all N under a category to open its tab.

Typing @counts finds the project file counts.csv as the top match

To jump straight to one category, type its letter and a space after @:

TypeShows
@f Files
@o Folders
@t Tools
@w Workflows
@p Pipelines
@d Databases
@a Atlas
@y Your Data

For example, @t boltz lists tools related to "boltz", with Boltz2 near the top, and @w scrna lists the single-cell workflows.

If a tab has no match, the picker says "No match for …" and offers Search all tabs. If you're looking for a file in another project, switch project first: the picker reminds you with "File in another project? Switch project to find it."

Preview a resource

Hover an item, or move to it with the arrow keys, to see its Resource Preview on the right: what it does, where it runs, its cost per run and, for tools and pipelines, the parameters you can set. Workflows also show their numbered steps. Before you hover anything, the pane shows Tips for the keyboard shortcuts.

Resource Preview for the Boltz2 tool, with its parameters

For a pipeline, the preview also shows its version and cost per hour.

Searching pipelines with @p rnaseq

Add one or several resources

  • One resource: click it, or move to it with the arrow keys and press Enter.
  • Several at once: Shift-click each item, or press Shift + Enter on each. The preview pane says how many are selected. Then press Enter or click Insert N to add them all.
  • Open a file in full: double-click it (or press Ctrl + Enter) to preview it instead of mentioning it.
  • Close the picker: press Esc.

Each mention becomes a pill in your message. Items you've already mentioned show "Already mentioned 1x", so you don't add them twice. To remove a mention, click the x on its pill.

Set a tool's or pipeline's parameters

You don't have to set anything: Drylab fills in a tool's options from its defaults and your message. To choose them yourself, click the tool's pill in your message to open its panel, fill in what you need and click Save parameters. Fields you leave empty are still picked for you ("Empty fields: the agent picks them from the tool's defaults and your message."). Reset puts the defaults back.

The Boltz2 parameter panel

Pipelines group their parameters into sections.

Choose files from a dataset or bucket

Some pills open a panel when you click them in your message:

  • Atlas dataset: shows the dataset's species, tissue, disease state and files. Tick the files you need, then click Use N files, or Use all files. View on Atlas opens the dataset's page.
  • Your Data bucket: browse the bucket's folders and files. Tick files or folders to send only those, then click Use N items, or Use whole bucket.

The footer of each panel says what will be sent ("Sends: …").

To run the analysis again on a different part of the bucket, click the bucket pill in a message you already sent. The panel opens with that message's selection. Change it, then click Analyze N items or Analyze entire bucket. Drylab adds a bucket pill with the new selection to your chatbox; type your request and send it. The earlier message stays as it was. The bucket must still be connected and mounted; otherwise you'll see "This bucket is no longer connected."

A bucket must be attached before you can mention it. If it isn't, you'll see "Mount this bucket first" with a link to Open Data Connector. See Data Connector.

Examples

  • @counts.csv load this and show the column names and shape
  • @counts.csv and @metadata.csv run DE of tumor vs normal paired by patient
  • @figure1.png what does this plot show and are there any issues?
  • @Boltz2 predict the complex of this protein with the ligand in @ligand.sdf
  • @scRNA QC + Clustering on my @pbmc.h5ad
  • @p rnaseq then "run this on the FASTQ files in @fastq_run_01 against the human reference"
  • Compare my clusters in @pbmc.h5ad with a matching dataset from @a pbmc

Use resources together

A typical chain for a single-cell project:

  1. @nf-core Scrnaseq to process raw reads into count matrices.
  2. @scRNA QC + Clustering to set QC cutoffs and pick a clustering resolution.
  3. @CellxGene to pull a reference atlas for annotation.
  4. @ChEMBL to look up drugs targeting the marker genes you found.

Not sure what to use? Ask: "I have scRNA-seq data. Which tools and databases should I use for annotation and differential expression?"

Tips

  • Files appear in the picker once they are uploaded to the current project. See Upload Files. In the Desktop App, the Files tab searches the files on your computer.
  • Choose the project first, so the Files and Folders tabs show the right files. See Projects.
  • Mention the data and the method in the same message: "@counts.csv with @DESeq2", not two separate messages.
  • The + button in the chatbox offers Add Vault files, folders or tools, which opens the same picker.
  • On a project page, you can also drag a file from Sources onto the chatbox to mention it.
  • To ask Drylab about a figure in the chat, use Mention on the figure card. See Edit Figures.

On this page