4.2.4 DenseLSTM

hcwang and qxdu edited on Aug 4, 2023, 1 version

DenseLSTM Model Architecture

DenseLSTM is a predictive model for designing flanking sequence in mammalian cells[1]. With this model, researchers have designed E. coli constitutive, isopropyl-beta-D-1-thiogalactopyranoside (IPTG)-inducible, and mammalian cell doxycycline (Dox)-inducible promoters with significant performance improvements.A schematic diagram of the whole workflow has been provided.

Here, in order to facilitate users' understanding of the process of the model in biological sequences, we provide a more detailed operational pipeline.

Caution: We recommend Input sequences should have a length exceeding 100bp, as we found that the two models related to densenet have limited performance on short sequences.

Input Parameters

All parameters should be defined during the initialization phase. We have encapsulated the source code, thus all predictive models have unified input and output parameters. There are two types of parameters, one should be defined during the initialization phase (Initialization), and the other should be defined during the training/sampling phase (Training/Predicting).

Initialization params

params	description	default value
batch_size	training batch size	64
length	sequential length of the training dataset	50
model_name	parameter that controls the saving path under "./checkpoints"	denselstm
epoch	training epochs	200
patience	earlystopping when the indicators no longer change	50
log_steps	logging the output/criterias of the model every print_epoch epochs	10
save_steps	saving the result of model every save_epoch epochs	20
exp_mode	the processing mode for expression input	log2

Training/Predicting params

params	description	default value	flexible stage
dataset	training dataset sequences path, fasta file	None	train()
labels	training dataset expression path, txt file, each line an expression corresponding to dataset	None	train()
savepath	final model saving path directory	None	train()
model_path	model loading directory	None	predict()/predict_input()
data_path	dataset to be predicted , fasta file	None	predict()
inputs	data for predict_input, can be datapath, sequence list or onehot encoded data	None	predict_input()
mode	input mode for predict_input, can be "path","data" or "onehot"	"path"	predict_input()

Caution: predict() function will directly generate samples in checkpoint path, but predict_input() will not generate the file automatically.

Demo

A demo for model training/predicting is described below:

from gpro.predictor.denselstm.denselstm import DenseLSTM_language
model = DenseLSTM_language(length=50, epoch=400, patience=10, exp_mode="direct")

# Train
default_root = "your working directory"
dataset = os.path.join(default_root, 'data/seq.txt')
labels  = os.path.join(default_root, 'data/exp.txt')
save_path = os.path.join(default_root, 'checkpoints/')
model.train(dataset=dataset,labels=labels,savepath=save_path)

# Predict
model_path = os.path.join(default_root, "checkpoints/denselstm/checkpoint.pth")
data_path =  os.path.join(default_root, "data/example.txt")
model.predict(model_path=model_path, data_path=data_path)

# Predict input
res = model.predict_input(model_path=model_path, inputs=data_path)
print(res)

Citations

[1] Zhang P, Wang H, Xu H, et al. Deep flanking sequence engineering for efficient promoter design[J]. bioRxiv, 2023: 2023.04. 14.536502.