Versions

Change¶

Version	Date
`0.0.0.0.1`
`0.0.1`

Before `0.0.1`¶

The version 0.0.0.0.1 of DeepdlncUD was used to infer a small molecule-lncRNA regulation type per instance only. This process is simplified as follows.

Small molecule¶

Clopidogrel (CID: 60606)

COC(=O)[C@H](C1=CC=CC=C1Cl)N2CCC3=C(C2)C=CS3

Target¶

HIF1A-AS1

>HIF1A-AS1
CGCCGCCGGCGCCCTCCATGGTGAATCGGTCCCCGCGATGTCTTCACGGCGGGCGGCCCCCAGGCTCGCTCCGGCCTAAGCGCTGGCTCCCTCCACACGCGGAGAAGAGAAGGAAAGACTACAGTTCAACTGTCAATTGGTTGATCACCCGGATTTTATCTACACCTTAGCCTATGGTTGTTCATCTCGTCTCTGCCTATGGCCCATTGACTCCCGGATCCCAGCTCCATTCTTCGGTACTTTACGCACCCTGCTTCCAGTACCCCAACCAGAAGAATATATATAGCAGTTAACTGTCAGCTGGCGAAAAGGAGGAAAATTCAGGAAGATAAAATAGCTGAATGAATTATCCCCGCTCCAGAACGCAGAGGAAAAATGAAATGGCCAGACCCAGATGTTAAAAATGTGTTCCTTGCTCTTTCCTGCCCTAGCAAGGGCTGTTCCATGTTTAGGGGATGAATGCCGCTGAGAGTATTAGCAAAAATACATGTGTCATTGAGTCTGAGGAAGATAACTGAGACATACAGGTATTTCTCATAATGCATGTGGGCATCCATAGACATATTCTTAAATGGCTTAAGGACTTGGAAACTACCTCTAGGAAACCTGAAACTTGAATGTTGGTCCACTAGGGAGAAGAAATGTTCCATTA

Inference¶

deepdlncud -m LSTMCNN -sm deepdlncud/data/example/60606.txt -lncr deepdlncud/data/example/HIF1A-AS1.fasta -mf /the/path/you/prefer/model/lstmcnn -o ./out.deepdlncud

Tensorflow¶

Major change in deep learing libraries

In Keras 3, the way to call models becomes completely different from that in Keras 2.

Table 1:The way to use models.

Item	TF Keras 2	TF Keras 3
model format	`.h5` or TF `SavedModel`	`.keras` (new) + `.h5` (legacy)
`subclass` Models	supported via `SavedModel`	`TFSMLayer` for loading

Comparison¶

Our experience suggests that changes in Python versions have a smaller impact compared to changes in TensorFlow versions.

Table 2:The way to use models.

Python	Tensorflow verisons	Supported versions
`3.11`	🌟`2.14`🌟, `2.15`, `2.16`	TF Keras 2, TF Keras 3
`3.12`	`2.17`, `2.18`, `2.19`	only TF Keras 3