Skip to main content
. 2019 Jul 5;35(14):i333–i342. doi: 10.1093/bioinformatics/btz337

Table 1.

Pearson correlation coefficients by subcellular fraction of various model and input settings. Numbers in bold are the maximum of their row

Full-length RNA Inputs
Fixed-length Inputs (4 kb)
5Mer Inputs
Dataset Compartment RNATrackerseq NoLSTM RNATrackerseq NoAttention Seq+Struct Seq×Struct DNN-5Mer NN-5Mer
CeFra-Seq Cytosol 0.705 0.676 0.685 0.625 0.666 0.652 0.637 0.558
Insoluble 0.641 0.626 0.619 0.557 0.604 0.591 0.552 0.478
Membrane 0.540 0.509 0.469 0.306 0.451 0.409 0.421 0.384
Nuclear 0.542 0.515 0.502 0.379 0.475 0.449 0.485 0.432
APEX-RIP ER 0.626 0.554 0.485 0.150 0.469 0.394 0.407 0.368
Mitocondria 0.482 0.449 0.423 0.139 0.376 0.320 0.292 0.224
Cytosol 0.561 0.522 0.501 0.259 0.493 0.423 0.446 0.363
Nuclear 0.456 0.402 0.397 0.235 0.384 0.338 0.332 0.238

Note: NoLSTM and NoAttention are the two ablation tests without the bidirectional LSTM or the attention module.