Those building predictive models from transcriptomic data are faced with two conflicting perspectives. The first, based on the inherent high dimensionality of biological systems, supposes that complex non-linear models such as neural networks will better match complex biological systems. The second, imagining that complex systems will still be well predicted by simple dividing lines prefers linear models that are easier to interpret. We compare multi-layer neural networks and logistic regression across multiple prediction tasks on GTEx and Recount3 datasets and find evidence in favor of both possibilities. We verified the presence of non-linear signal when predicting tissue and metadata sex labels from expression data by removing the predictive linear signal with Limma, and showed the removal ablated the performance of linear methods but not non-linear ones. However, we also found that the presence of non-linear signal was not necessarily sufficient for neural networks to outperform logistic regression. Our results demonstrate that while multi-layer neural networks may be useful for making predictions from gene expression data, including a linear baseline model is critical because while biological systems are high-dimensional, effective dividing lines for predictive models may not be.

Motivation: Classification by supervised machine learning greatly facilitates the annotation of protein characteristics from their primary sequence. However, the feature generation step in this process requires detailed knowledge of attributes used to classify the proteins. Lack of this knowledge risks the selection of irrelevant features, resulting in a faulty model. In this study, we introduce a supervised protein classification method with a novel means of automating the work-intensive feature generation step via a Natural Language Processing (NLP)-dependent model, using a modified combination of n-grams and skip-grams (m-NGSG). Results: A meta-comparison of cross-validation accuracy with twelve training datasets from nine different published studies demonstrates a consistent increase in accuracy of m-NGSG when compared to contemporary classification and feature generation models. We expect this model to accelerate the classification of proteins from primary sequence data and increase the accessibility of protein characteristic prediction to a broader range of scientists. Availability and implementation: m-NGSG is freely available at Bitbucket: https://bitbucket.org/sm_islam/mngsg/src. A web server is available at watson.ecs.baylor.edu/ngsg. Contact: [email protected]. Supplementary information: Supplementary data are available at Bioinformatics online.