Organ |
organ from which the cluster was derived |
Cluster_ChromBPNet |
cluster ID in the format used for ChromBPNet training & analyses (as in Table S2) |
L1_annot |
level 1 cell type annotation (n=203) |
Total_n_fragments |
total number of fragments for all cells in the cluster |
Model |
ID for the model (defined by cluster and fold) |
Fold |
training fold (0 to 4) |
Pass_QC |
binary variable indicating whether the model passed performance thresholds and QC and was used for downstream analysis |
Reason |
reason for rejecting model |
counts_metrics.peaks.spearmanr |
ChromBPNet model performance metric: Spearman's rank correlation between observed and predicted log counts in peaks on test chromosomes |
counts_metrics.peaks.pearsonr |
ChromBPNet model performance metric: Pearson correlation between observed and predicted log counts in peaks on test chromosomes |
counts_metrics.peaks.mse |
ChromBPNet model performance metric: mean squared error between observed and predicted log counts in peak on test chromosomes |
profile_metrics.peaks.median_jsd |
ChromBPNet model performance metric: median Jensen Shannon Divergence between observed and predicted profiles for peaks on test chromosomes (lower is better) |
profile_metrics.peaks.median_norm_jsd |
ChromBPNet model performance metric: median min-max normalized JSD (higher is better) |
tn5_motif_1 |
ChromBPNet model performance metric: max of predicted marginal footprint for Tn5 motif #1 in bias-corrected model (high response indicate that the model is still senstive to Tn5 sequence bias) |
tn5_motif_2 |
as above for Tn5 motif #2 |
tn5_motif_3 |
as above for Tn5 motif #3 |
tn5_motif_4 |
as above for Tn5 motif #4 |
tn5_motif_5 |
as above for Tn5 motif #5 |
max_tn5_response |
max of tn5motif1, tn5motif2, tn5motif3, tn5motif4, tn5motif5 |