Import data

The first screen has one button: Import file…. It offers .h5ad, .rds, .rda, .rdata, and .csv.

Everything else in the wizard follows from what you load here.

Cell templates

This list is the template library every import starts from. Add files… puts a file on it; Remove file… takes one off. Your host's own files seed it, as ordinary entries — remove one you do not want and it is gone for good.

The files are only named here; they are read at the cell-templates step, so a file that cannot be read is reported there.

What BIWT reads from each format

Format	Read via	Supported objects
.h5ad	anndata.read_h5ad	AnnData
.rds	rpy2 + anndata2ri	Seurat, SingleCellExperiment, SpatialExperiment, SummarizedExperiment
.rda / .rdata	base::load() in R	Same, as the first object in the workspace
.csv	pandas.read_csv	A flat table, one row per cell

Whatever the source, BIWT normalizes it to the same internal shape: a table of per-cell metadata (obs), an optional array of spatial coordinates (obsm), and a cell count.

.rda / .rdata files with several objects

BIWT takes the first object in the workspace. If you saved several objects together, re-save just the one you want.

How spatial coordinates are found

BIWT looks for coordinates in two places, in order:

1. An obsm spatial array — the normal location for AnnData and SCE objects.
2. Columns in obs — checked in priority order: x/y/z-style names first, then, as a last resort, the 10x Visium pixel columns imagecol and imagerow.

The Visium fallback needs one adjustment: image rows increase downward, so imagerow is flipped (y = rowmax − imagerow) to give a conventional y-up system. imagecol maps to x directly.

When coordinates come from obs columns rather than an obsm array, BIWT synthesizes obsm["spatial"] from them so the scatter plots on later screens have something to draw.

Coordinates found this way are in data units

BIWT reports the extent in a generic data unit and infers no unit name from your file. Visium pixel coordinates in particular are not microns. Converting them is the job of the scale factor in the domain editor — and Visium is the one case where BIWT can read the factor (µm per pixel) out of the file and pre-fill it for you.

Probability columns

If your table has columns ending in _probability — for example T_cell_probability, Tumor_probability — BIWT recognizes them as per-spot cell-type probabilities. Combined with spatial coordinates, that unlocks the spot deconvolution step.

Re-importing

Importing a second file resets the whole session. Every choice you made about the previous file — cluster column, merges, renames, counts, positions — is discarded. There is no partial carry-over.

When import fails

The error dialog tells you what failed and why. Failures fall into two groups:

Something is wrong with the file. Unsupported extension, malformed CSV, an R object of a class BIWT does not handle, an empty R workspace. Fix the file.

Something is wrong with your environment. A missing optional dependency, or an R stack that cannot load your object. These dialogs carry a link to the setup instructions, because that is where the fix is:

• anndata is required for .h5ad files → install biwt[anndata]
• rpy2 and anndata2ri are required for R files → install biwt[seurat] and an R stack (recipe)
• anndata2ri activation failed → you have anndata2ri 2.0+; pin it below 2
• Failed to read … as R object → usually SeuratObject missing from the R that rpy2 bound to, or an ABI-mismatched R (troubleshooting)

A failed import leaves you on the import screen with nothing loaded, so you can fix the problem and try again without restarting.

Next

Spot deconvolution → if your data has probability columns and coordinates; otherwise cluster column →.