SIMON: Open-Source Knowledge Discovery Platform

Abstract

Data analysis and knowledge discovery has become more and more important in biology and medicine with the increasing complexity of biological datasets, but the necessarily sophisticated programming skills and in-depth understanding of algorithms needed pose barriers to most biologists and clinicians to perform such research. We have developed a modular open-source software, SIMON, to facilitate the application of 180+ state-of-the-art machine-learning algorithms to high-dimensional biomedical data. With an easy-to-use graphical user interface, standardized pipelines, and automated approach for machine learning and other statistical analysis methods, SIMON helps to identify optimal algorithms and provides a resource that empowers non-technical and technical researchers to identify crucial patterns in biomedical data.

Keywords: artificial intelligence; autoML; bioinformatics; computational biology; data mining; data science; machine learning; software; systems biology.

Graphical abstract

Figure 1

SIMON Machine Learning Workflow Step 1. Building predictive models. (A) Screenshot of the SIMON graphical user interface demonstrating input selection for machine learning analysis, such as predictors and response (outcome) variables, additional exploration classes, training/test split, pre-processing functions, and desired machine learning algorithms. Step 2. Model evaluation and selection. Comparison of (B) box plots of performance measurements calculated for 11 predictive models and (C) receiver operating characteristic (ROC) curves built on the SISA dataset. Each boxplot shows the distribution of data as minimum (Q1−1.5×IQR), first quartile (Q1), median (Q2), third quartile (Q3), and maximum (Q3+1.5×IQR). Data outside of minimum and maximum values (outliers) are shown as circles. IQR, interquartile range.Comparison of ROC curves calculated from the training (average value calculated using 10-fold cross-validation repeated three times) and test sets on (D) datasets with missing values (Cyclists and VAST) and (E) high-dimensional datasets (Zeller and LIHC). Step 3. Feature selection. (F) The variable importance score table for each feature and graphical visualization of the selected features from the Cyclists dataset. Step 4. Exploratory analysis. (G) Correlation analysis on the Cyclists dataset. (H) Clustering analysis on the VAST dataset.