File size: 2,339 Bytes
70070cd | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 | # Quick Start
!!! note "Availability"
Model weights and inference code are coming soon. The examples below reflect the planned API interface.
## Load a Pretrained Model
```python
from xcell import XCell
model = XCell.from_pretrained("Xaira-Therapeutics/X-Cell", variant="mini")
```
`variant="mini"` loads X-Cell Mini (55M parameters).
---
## Predict from an AnnData Object
```python
import anndata as ad
from xcell import XCell
model = XCell.from_pretrained("Xaira-Therapeutics/X-Cell", variant="mini")
# Load your control cells
adata = ad.read_h5ad("control_cells.h5ad")
# Predict transcriptional response to a single-gene CRISPRi knockdown
predictions = model.predict(adata, perturbation="BRCA1")
# predictions is an AnnData with predicted perturbed expression in .X
print(predictions)
```
!!! note "Input expectations"
`adata` should contain log-normalized (log1p CP10k) expression values.
X-Cell covers all ~19,000 protein-coding genes; genes not in the vocabulary are zero-imputed.
---
## Predict from `.h5ad` File Paths
```python
# Single file path
predictions = model.predict("control_cells.h5ad", perturbation="BRCA1")
# Multiple files — cells are pooled across datasets
predictions = model.predict(
["screen1.h5ad", "screen2.h5ad"],
perturbation="BRCA1",
)
```
---
## Batch Predict Across Perturbations
```python
import pandas as pd
perturbations = ["BRCA1", "TP53", "MYC", "EGFR"]
results = {}
for pert in perturbations:
results[pert] = model.predict(adata, perturbation=pert)
# Concatenate into a single AnnData
import anndata as ad
all_predictions = ad.concat(results, label="perturbation")
```
---
## Interpreting Outputs
The returned `AnnData` contains:
| Field | Description |
|-------|-------------|
| `.X` | Predicted perturbed expression (log1p CP10k, same shape as input) |
| `.obs["perturbation"]` | Perturbation name |
| `.var` | Gene metadata (same as input `adata.var`) |
To compute predicted fold-change relative to control:
```python
import numpy as np
# Pseudobulk fold-change
ctrl_mean = adata.X.mean(axis=0)
pred_mean = predictions.X.mean(axis=0)
fold_change = pred_mean - ctrl_mean # in log space, this is log fold-change
# Top differentially expressed genes
top_genes = np.argsort(np.abs(fold_change))[::-1][:20]
print(adata.var_names[top_genes])
```
|