----cells_by_motifs.txt: contains aggregated motif scores for cells and has following columns:

filter: filter name that their PWM was used to scan the accessible sites using TomTom 
cell: cell barcode
motif score: the total number of sites that had this matched this PWM
total motifs: total number of motifs detected in all the accessible sites for this cell
motif activity: shows the relative frequency of this motif relative with all the motifs scored for this cell (motif score/total motifs)

----cells_by_motifs.rds: the same as cells_by_motifs.txt but compressed to be loaded easily in R using readRDS.

----filter_sd_mean.txt: contains information content (ic), mean, and standard deviation(std) for the filter activity over the test data.

----filter_influences.txt: contains filter influence (infl) for each filter for each cluster (cluster and subset_cluster)

----filters_meme.txt: contains PWMs for all the first layer filters in the trained CNN

----train_test_data.h5: train and test data used to train and test the CNN in the hdf5 format

----tomtom_hits.txt: contains filters matched to known motifs using TomTom with value < 0.1 and has following columns:
#Query ID: filter name 
Target ID: motif name in Hocomoco database
other columns are the default outputs in meme and are explained here http://meme-suite.org/doc/tomtom.html?man_type=web:

Optimal offset: the offset between the query and the target motif
Overlap: the number of positions of overlap between the two motifs.
Query consensus sequence.
Target consensus sequence.
Orientation: Orientation of target motif with respect to query motif.

----training_params.txt: parameters used to train the CNN using Basset framework

----trained_model.th: CNN model trained using Basset framework