---
license: cc-by-4.0
pretty_name: "spaCR example: optical pooled screen"
tags:
- microscopy
- optical-pooled-screen
- crispr
- in-situ-sequencing
- spacr
size_categories:
- n<1K
source_datasets:
- "BioImage Archive S-BIAD394"
---

# spaCR example: optical pooled screen

Two neighbouring 10X fields (sites 331 and 332, the last two of the
333-field well) of the in-situ sequencing acquisition, with all eleven
sequencing cycles: `sequencing/c1/` holds cycle 1 as one five-plane stack
(DAPI, CY3, A594, CY5, CY7) and `sequencing/c2/` ... `c11/` hold each later
cycle's four base channels as separate files (1,480 x 1,480 px, uint16).
`library/pool10_prefixes.csv` is the screen's guide library as 11-base
barcode prefixes. About 394 MB.

Sites 331 and 332 are chosen because spaCR's OPS engine takes a well's field
count from its highest site number, and the last field keeps the measured
333-field layout. There are no phenotype images: a 20X phenotype field is
280 MB on its own, and placing phenotype fields needs anchors spread over the
whole well.

## Source and licence

A small, **unmodified** sample of the primary fixed-cell optical pooled screen
of Funk *et al.* 2022, *The phenotypic landscape of essential human genes*,
Cell 185(24):4634-4653 ([doi:10.1016/j.cell.2022.10.017](https://doi.org/10.1016/j.cell.2022.10.017),
PMID 36347254), deposited in the BioImage Archive as
[S-BIAD394](https://www.ebi.ac.uk/biostudies/BioImages/studies/S-BIAD394):
plate `20200202_6W-LaC024A`, well A1. Every image is byte-identical to the
archived file; `manifest.csv` inside the archive gives each file's size,
SHA-256 and its path under
`https://ftp.ebi.ac.uk/biostudies/fire/S-BIAD/394/S-BIAD394/Files/`.

The BioImage Archive releases directly submitted data under CC0 or CC-BY-4.0
([policy](https://www.ebi.ac.uk/bioimage-archive/help-policies/)); S-BIAD394
carries no other licence attribute. This sample is redistributed under
CC-BY-4.0 with attribution to the original authors. **Cite the paper above**
if you use it.

The guide library is derived from `pool10_design.csv` in the authors' [code repository](https://github.com/lukebfunk/OpticalPooledScreens) (MIT licence) exactly as their Snakefile derives it.

## Use in spaCR

In the Mask screen press **OPS**, then **Load test data...** under
*OPS input*: the set is downloaded into
`~/.cache/spacr/example_data/ops_screen/` and `genotype_source`, `dst_root`,
`ops_library` and `plate` are filled in. **Run** stitches the two fields,
segments the nuclei with Cellpose and decodes each nucleus's barcode against
the library; the tables go to `ops_output/measurements.db`. From Python:

```python
from spacr.ops_engine import run_ops
run_ops({"genotype_source": "<folder>/sequencing",
         "dst_root": "<folder>/ops_output",
         "ops_library": "<folder>/library/pool10_prefixes.csv",
         "plate": "20200202_6W-LaC024A"})
```

Measured on this sample with spaCR's CPU path (`ops_gpu` off): the stitch
places both fields on one accepted edge; Cellpose (cpsam) numbers 10,675
nuclei; decoding finds 22,733 spots over eleven cycles, 80 % of them an exact
match to a library prefix, and assigns a barcode to 3,729 nuclei, 3,549 of
them library-exact (95 %). On a CPU the segmentation takes about an hour; on a
GPU it takes minutes. The decode takes about 20 s.

## Files

One uncompressed tar, `spacr-example-ops.tar`, unpacked with Python's tar data filter by
spaCR. `SHA256SUMS` holds the archive's checksum.