Compare with Public Data

To check your results against a published reference, mention a public dataset from Drylab Atlas in your chat. Pick only the files you need, and Drylab reads them where they're stored, next to your data. Use it for a QC sanity check, a cell-type comparison or a mapping to a reference atlas.

When to use it

• You want to know whether your sample's QC numbers (cells recovered, genes per cell) are in a normal range for the same tissue and chemistry.
• You want to check your clusters or cell-type labels against a published dataset of the same tissue.
• You need a reference to map or annotate your cells, without downloading a large public dataset first.
• You want to compare your bulk or single-cell results with a disease or healthy dataset from the literature.

Prompt examples

Type @a and a few words to find a dataset, then pick it. It becomes a chip in your message.

QC sanity check:

My PBMC sample has about 8,500 cells with a median of 1,900 genes per cell. Compare these QC numbers with the public dataset @a pbmc and tell me in a short table whether my sample is in the same range.

Cell types against a reference:

Annotate the clusters in @pbmc_sample.h5ad, then compare the cell-type proportions with the public dataset @a pbmc. Show both side by side in one bar chart.

Map to a reference:

Use a healthy-donor dataset from @a pbmc as a reference and transfer its cell-type labels to the cells in @pbmc_sample.h5ad. Report how confident each label is and list cells that don't match well.

Tissue-level comparison:

Compare the marker genes of my tumor clusters in @tumor_clusters.csv with the public dataset @a liver. Which of my clusters look like cell types in the reference, and which look new?

Steps

1. In the chatbox, start your request, then type @a (the letter a and a space) and a few words, such as @a pbmc. The Atlas tab lists matching datasets.
2. Hover a dataset to see its Resource Preview: product, software, species, anatomical entity and disease state. Pick the one that matches your experiment best, then click it and click Click to use resource. A chip that reads All files appears in your message.
3. Click the chip to choose its files. Untick what you don't need, or click Clear and tick only the files you want. The header shows the count, for example "2 of 15 selected". Click Use 2 files (or Use all files). The chip now reads 2 files.
4. Mention your own data too, for example @pbmc_sample.h5ad, and finish your request.
5. Send it. Drylab reads the public files where they're stored, compares them with your data and answers. The dataset appears under Datasets in the chat's file list, with the number of files you chose.

The steps are the same in the Desktop App. There, Drylab downloads only the files you chose to your computer when it needs them.

Tips

• Search with one or two words. @a pbmc lists many PBMC datasets; a long phrase can match nothing. To browse and filter the whole catalog instead, use Drylab Data and click Try in Drylab.
• Pick the summary files for a quick check. For QC comparisons, a dataset's summary file (for example …_count_summary.json) is often enough. Drylab reads it in seconds instead of opening the full count matrix.
• Match the experiment. Compare like with like: same species, tissue and chemistry (for example Universal 5' vs 3'). The Resource Preview shows these before you pick.
• Change the files before you send. Once a message is sent, its dataset chip is read-only. To use other files, mention the dataset again in a new message.

Related

• Drylab Atlas
• Choose which files to use
• Mention Files and Resources
• Analyze Data in Your Own Bucket