ENNGene: an Easy Neural Network model building tool for Genomics

Abstract

Background: The recent big data revolution in Genomics, coupled with the emergence of Deep Learning as a set of powerful machine learning methods, has shifted the standard practices of machine learning for Genomics. Even though Deep Learning methods such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) are becoming widespread in Genomics, developing and training such models is outside the ability of most researchers in the field.

Results: Here we present ENNGene-Easy Neural Network model building tool for Genomics. This tool simplifies training of custom CNN or hybrid CNN-RNN models on genomic data via an easy-to-use Graphical User Interface. ENNGene allows multiple input branches, including sequence, evolutionary conservation, and secondary structure, and performs all the necessary preprocessing steps, allowing simple input such as genomic coordinates. The network architecture is selected and fully customized by the user, from the number and types of the layers to each layer's precise set-up. ENNGene then deals with all steps of training and evaluation of the model, exporting valuable metrics such as multi-class ROC and precision-recall curve plots or TensorBoard log files. To facilitate interpretation of the predicted results, we deploy Integrated Gradients, providing the user with a graphical representation of an attribution level of each input position. To showcase the usage of ENNGene, we train multiple models on the RBP24 dataset, quickly reaching the state of the art while improving the performance on more than half of the proteins by including the evolutionary conservation score and tuning the network per protein.

Conclusions: As the role of DL in big data analysis in the near future is indisputable, it is important to make it available for a broader range of researchers. We believe that an easy-to-use tool such as ENNGene can allow Genomics researchers without a background in Computational Sciences to harness the power of DL to gain better insights into and extract important information from the large amounts of data available in the field.

Keywords: Convolutional Neural Network; Deep Learning; Evolutionary Conservation Score; GUI; RNA Secondary Structure; Recurrent Neural Network.

Fig. 1

a The Graphical User Interface (GUI) of ENNGene—ENNGene is fully operated via the GUI. Users define the input parameters using simple interactive elements, such as dropdown menus or checkboxes. Warnings and hints are displayed via the GUI in a user-friendly way directly as the user interacts with it. Web browser being at the basis of the GUI, interactive plots or results are visualized immediately throughout or after the calculations. b Simplified data flow—ENNGene comprises multiple subsequent modules with separate functionality, covering the whole process from input preparation and network architecture definition to model evaluation and interpretation

Fig. 2

a Precision-recall curve—the precision-recall metric indicates the relationship between the model’s positive predictive value (precision) and sensitivity (recall) at various thresholds. b Receiver Operating Characteristic (ROC) curve—the ROC metric is calculated as a ratio between the true positive rate and the false positive rate at various thresholds. Both the metrics, precision-recall and ROC calculated by ENNGene, are adjusted for multi-class classification problems and thus can be applied to models with any number of classes. Both curves and other metrics (accuracy, loss, AUROC) are a standard part of exported results after a model evaluation, optionally with Integrated Gradients’ scores. c Integrated Gradients visualization—IG scores of ten sequences with the highest predicted score per class are directly visualized in the browser. Scores are displayed in separate rows for each input type used—sequence, secondary structure, and conservation score. The higher the nucleotide’s attribution to the prediction of a given class, the more pronounced is its red color. On the other hand, the blue color means a low level of attribution

Fig. 3

Simplified representation of model architecture. The model in this example was trained on 150 nt long sequences using all three available input types—sequence, secondary structure, and conservation score—each represented by a separate model branch. After the network extracts information from the separate inputs, the branches are concatenated, and the network continues learning interdependencies by looking at the combined information via dense or recurrent layers. Boxes represent individual layers, while the adjacent numbers indicate the data dimensionality. A plain graphical representation of the network architecture is produced and exported by ENNGene for every trained model