Lin Chih-Hsu, Lichtarge Olivier
Department of Molecular and Human Genetics.
Department of Biochemistry and Molecular Biology.
Bioinformatics. 2021 Sep 9;37(17):2675-2681. doi: 10.1093/bioinformatics/btab137.
Cancer dependencies provide potential drug targets. Unfortunately, dependencies differ among cancers and even individuals. To this end, visible neural networks (VNNs) are promising due to robust performance and the interpretability required for the biomedical field.
We design Biological visible neural network (BioVNN) using pathway knowledge to predict cancer dependencies. Despite having fewer parameters, BioVNN marginally outperforms traditional neural networks (NNs) and converges faster. BioVNN also outperforms an NN based on randomized pathways. More importantly, dependency predictions can be explained by correlating with the neuron output states of relevant pathways, which suggest dependency mechanisms. In feature importance analysis, BioVNN recapitulates known reaction partners and proposes new ones. Such robust and interpretable VNNs may facilitate the understanding of cancer dependency and the development of targeted therapies.
Code and data are available at https://github.com/LichtargeLab/BioVNN.
Supplementary data are available at Bioinformatics online.
癌症依赖性提供了潜在的药物靶点。不幸的是，不同癌症甚至个体之间的依赖性存在差异。为此，可见神经网络（VNN）因其强大的性能和生物医学领域所需的可解释性而颇具前景。
我们利用通路知识设计了生物可见神经网络（BioVNN）来预测癌症依赖性。尽管参数较少，但BioVNN的性能略优于传统神经网络（NN），且收敛速度更快。BioVNN也优于基于随机通路的神经网络。更重要的是，通过与相关通路的神经元输出状态相关联，可以解释依赖性预测，这揭示了依赖性机制。在特征重要性分析中，BioVNN重现了已知的反应伙伴并提出了新的伙伴。这种强大且可解释的VNN可能有助于理解癌症依赖性并推动靶向治疗的发展。
代码和数据可在https://github.com/LichtargeLab/BioVNN获取。
补充数据可在《生物信息学》在线获取。
