Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic Mutations
Several challenges appear in the application of deep learning to genomic data. First, the dimensionality of input can be orders of magnitude greater than the number of samples, forcing the model to be prone to overfitting the training dataset. Second, each input variable’s contribution to the predic...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-08-01
|
| Series: | Biomolecules |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2218-273X/10/9/1249 |
| _version_ | 1827707360004538368 |
|---|---|
| author | Kazuma Kobayashi Amina Bolatkan Shuichiro Shiina Ryuji Hamamoto |
| author_facet | Kazuma Kobayashi Amina Bolatkan Shuichiro Shiina Ryuji Hamamoto |
| author_sort | Kazuma Kobayashi |
| collection | DOAJ |
| description | Several challenges appear in the application of deep learning to genomic data. First, the dimensionality of input can be orders of magnitude greater than the number of samples, forcing the model to be prone to overfitting the training dataset. Second, each input variable’s contribution to the prediction is usually difficult to interpret, owing to multiple nonlinear operations. Third, genetic data features sometimes have no innate structure. To alleviate these problems, we propose a modification to Diet Networks by adding element-wise input scaling. The original Diet Networks concept can considerably reduce the number of parameters of the fully-connected layers by taking the transposed data matrix as an input to its auxiliary network. The efficacy of the proposed architecture was evaluated on a binary classification task for lung cancer histology, that is, adenocarcinoma or squamous cell carcinoma, from a somatic mutation profile. The dataset consisted of 950 cases, and 5-fold cross-validation was performed for evaluating the model performance. The model achieved a prediction accuracy of around 80% and showed that our modification markedly stabilized the learning process. Also, latent representations acquired inside the model allowed us to interpret the relationship between somatic mutation sites for the prediction. |
| first_indexed | 2024-03-10T16:44:36Z |
| format | Article |
| id | doaj.art-206b1c8352e345b78117c7e86d330ac4 |
| institution | Directory Open Access Journal |
| issn | 2218-273X |
| language | English |
| last_indexed | 2024-03-10T16:44:36Z |
| publishDate | 2020-08-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Biomolecules |
| spelling | doaj.art-206b1c8352e345b78117c7e86d330ac42023-11-20T11:41:02ZengMDPI AGBiomolecules2218-273X2020-08-01109124910.3390/biom10091249Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic MutationsKazuma Kobayashi0Amina Bolatkan1Shuichiro Shiina2Ryuji Hamamoto3Division of Molecular Modification and Cancer Biology, National Cancer Center Research Institute, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045, JapanDivision of Molecular Modification and Cancer Biology, National Cancer Center Research Institute, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045, JapanDepartment of Diagnostic Imaging and Interventional Oncology, Graduate School of Medicine, Juntendo University, 2-1-1 Hongo, Bunkyo-ku, Tokyo 113-8421, JapanDivision of Molecular Modification and Cancer Biology, National Cancer Center Research Institute, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045, JapanSeveral challenges appear in the application of deep learning to genomic data. First, the dimensionality of input can be orders of magnitude greater than the number of samples, forcing the model to be prone to overfitting the training dataset. Second, each input variable’s contribution to the prediction is usually difficult to interpret, owing to multiple nonlinear operations. Third, genetic data features sometimes have no innate structure. To alleviate these problems, we propose a modification to Diet Networks by adding element-wise input scaling. The original Diet Networks concept can considerably reduce the number of parameters of the fully-connected layers by taking the transposed data matrix as an input to its auxiliary network. The efficacy of the proposed architecture was evaluated on a binary classification task for lung cancer histology, that is, adenocarcinoma or squamous cell carcinoma, from a somatic mutation profile. The dataset consisted of 950 cases, and 5-fold cross-validation was performed for evaluating the model performance. The model achieved a prediction accuracy of around 80% and showed that our modification markedly stabilized the learning process. Also, latent representations acquired inside the model allowed us to interpret the relationship between somatic mutation sites for the prediction.https://www.mdpi.com/2218-273X/10/9/1249deep learningDiet Networkslung cancerinterpretable neural networks |
| spellingShingle | Kazuma Kobayashi Amina Bolatkan Shuichiro Shiina Ryuji Hamamoto Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic Mutations Biomolecules deep learning Diet Networks lung cancer interpretable neural networks |
| title | Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic Mutations |
| title_full | Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic Mutations |
| title_fullStr | Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic Mutations |
| title_full_unstemmed | Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic Mutations |
| title_short | Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic Mutations |
| title_sort | fully connected neural networks with reduced parameterization for predicting histological types of lung cancer from somatic mutations |
| topic | deep learning Diet Networks lung cancer interpretable neural networks |
| url | https://www.mdpi.com/2218-273X/10/9/1249 |
| work_keys_str_mv | AT kazumakobayashi fullyconnectedneuralnetworkswithreducedparameterizationforpredictinghistologicaltypesoflungcancerfromsomaticmutations AT aminabolatkan fullyconnectedneuralnetworkswithreducedparameterizationforpredictinghistologicaltypesoflungcancerfromsomaticmutations AT shuichiroshiina fullyconnectedneuralnetworkswithreducedparameterizationforpredictinghistologicaltypesoflungcancerfromsomaticmutations AT ryujihamamoto fullyconnectedneuralnetworkswithreducedparameterizationforpredictinghistologicaltypesoflungcancerfromsomaticmutations |