PredSynRNA

In this study, we developed a new machine learning method, PredSynRNA, to predict the synaptic localization of human RNAs. The PredSynRNA models utilized developmental brain gene expression data as features to achieve relatively high performance during cross-validations and on an independent test dataset. PredSynRNA models were then applied to the prediction and prioritization of candidate RNAs localized to human synapses.
For details of this work, users can refer to our paper "Prediction of Synaptically Localized RNAs in Human Neurons Using Developmental Brain Gene Expression Data " published in Genes (Wei and Wang, 2022).

Requirements

Python 3
argparse
pandas
numpy
sklearn
keras
tensorflow
joblib

Prediction and prioritization of candidate human RNAs localized to synapses

The mRNA_prediction_data.csv contains over 7,000 brain-expressed RNAs and the lncRNA_prediction_data.csv contains over 3,000 brain-expressed lncRNAs which were not included in the training set. To predict and prioritize candidate human RNAs localized to synapses, the SVM, RF, and ANN models will be loaded. For example, for the prediction and prioritization of candidate human mRNAs localized to synapses, the following command line can be used:

python prediction.py -t data/training_dataset.csv -f data/mRNA_prediction_data.csv -s SVM.model -r RF.model -a ANN.hdf5 -p <path>

An output file (.xlsx) that includes the RNAs predicted as positive (dendritically localized) by each model with computed probability scores will be generated. The prediction results are stored in separate sheets. For example, in the sheet of "svm_proba_positive":

gene_id	name	predict_proba
ENSG00000244734	HBB	0.982107707700408
ENSG00000166426	CRABP1	0.973952101206992
ENSG00000051523	CYBA	0.969843064843169