Abstract
Gene expression profiling was made more cost-effective by the NIH LINCS program that profiles only ∼1, 000 selected landmark genes and uses them to reconstruct the whole profile. The D-GEX method employs neural networks to infer the entire profile. However, the original D-GEX can be significantly improved. We propose a novel transformative adaptive activation function that improves the gene expression inference even further and which generalizes several existing adaptive activation functions. Our improved neural network achieves an average mean absolute error of 0.1340, which is a significant improvement over our reimplementation of the original D-GEX, which achieves an average mean absolute error of 0.1637. The proposed transformative adaptive function enables a significantly more accurate reconstruction of the full gene expression profiles with only a small increase in the complexity of the model and its training procedure compared to other methods.
Highlights
Gene expression profiling is a great tool for medical diagnosis and deepening of the understanding of a disease (e.g., [1,2,3,4])
The goal of this and the following experiments is to establish the improvement as a result of using the novel transformative adaptive activation function (TAAF) in models trained on the full dataset
We compare the original D– GEX architectures equipped with the hyperbolic tangent activation function to architectures equipped with the novel TAAF with hyperbolic tangent as the inner activation function
Summary
Gene expression profiling is a great tool for medical diagnosis and deepening of the understanding of a disease (e.g., [1,2,3,4]). Despite a significant price drop in recent years, gene expression profiling is still too expensive for running large scale experiments. The original profiling method used linear regression for the profile reconstruction due to its simplicity and scalability; it was improved using a deep learning method for gene expression inference called D–GEX [6] which allows for a reconstruction of non-linear patterns. The D–GEX is a family of several similar neural networks with varying complexity in terms of the number of parameters used for the gene expression inference. Our main contribution is that the D–GEX family performance can be further improved by using more suitable activation functions even while keeping the architecture unchanged
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
