Classify Gramian Angular Field matrices with a pretrained CNN

Assigns GAF representations of signal chunks to one of four nucleotide categories (A, C, G, U) using a pretrained convolutional neural network. The model is loaded via load_keras_model and inference is performed through the TensorFlow/Keras backend.

Usage

predict_gaf_classes(gaf_list)

Arguments

gaf_list

List of GAF arrays (100, 100, 2), as produced by create_gaf_list.

Value

A named list with two elements:

chunkname

Character vector. Names of the classified signal chunks (format: <readname>_<index>)

prediction

Integer vector. Predicted nucleotide class for each chunk (0 = A, 1 = C, 2 = G, 3 = U)

Details

The function reshapes the input GAF list into a 4-D tensor (n_chunks x 100 x 100 x 2) and applies the pretrained CNN to predict the nucleotide class for each chunk. Predictions are returned as integer codes: 0 = A, 1 = C, 2 = G, 3 = U.

See also

create_gaf_list for preparing the input, load_keras_model for model loading, create_outputs for integrating predictions into final output

Examples

if (FALSE) { # \dontrun{

pl <- ninetails::predict_gaf_classes(gl)

} # }