eugene.interpret.motif_distance_dependence_gia

eugene.interpret.motif_distance_dependence_gia(model, sdata, feature_A, feature_B, tile_step=1, style='deAlmeida22', seq_var='seq', results_var='cooperativity', distance_var='distance', device='cpu', batch_size=128)

Calculate the dependence of the model predictions on the distance between two motifs.

Currently only supports a DeepSTARR style analysis in which a single motif is implanted into the center of a set of background sequences, and a second motif is tiled across the background at different distances from the first motif. A cooperativity score is calculated for the two motifs to quantify the dependence of the model predictions on the distance between the motifs.

Parameters:

• model (torch.nn.Module) – The model to use for predictions.

• sdata (xr.Dataset) – The dataset containing the background sequences.

• feature_A (str) – The first motif to implant in the center of the background sequences.

• feature_B (str) – The second motif to tile across the background sequences.

• tile_step (int, optional) – The step size to use to tile feature_B across the background sequences, by default 1.

• style (str, optional) – The style of the analysis, by default “deAlmeida22”.

• seq_var (str, optional) – The key for the sequence data in the dataset, by default “seq”.

• results_var (str, optional) – The key to store the results in the dataset, by default “cooperativity”.

• distance_var (str, optional) – The key to store the distances in the dataset, by default “distance”.

• device (str, optional) – The device to use for predictions, by default “cpu”.

• batch_size (int, optional) – The batch size to use for predictions, by default 128.