Enhanced Integrated Gradients: improving interpretability of deep learning models using splicing codes as a case study
Abstract Despite the success and fast adaptation of deep learning models in biomedical domains, their lack of interpretability remains an issue. Here, we introduce Enhanced Integrated Gradients (EIG), a method to identify significant features associated with a specific prediction task. Using RNA spl...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2020-06-01
|
| Series: | Genome Biology |
| Subjects: | |
| Online Access: | http://link.springer.com/article/10.1186/s13059-020-02055-7 |
| id |
doaj-c536fe96c96c49fa83f8088859105757 |
|---|---|
| record_format |
Article |
| spelling |
doaj-c536fe96c96c49fa83f80888591057572020-11-25T03:54:41ZengBMCGenome Biology1474-760X2020-06-0121112210.1186/s13059-020-02055-7Enhanced Integrated Gradients: improving interpretability of deep learning models using splicing codes as a case studyAnupama Jha0Joseph K. Aicher1Matthew R. Gazzara2Deependra Singh3Yoseph Barash4Department of Computer and Information Science, School of Engineering and Applied Science, University of PennsylvaniaDepartment of Genetics, Perelman School of Medicine, University of PennsylvaniaDepartment of Genetics, Perelman School of Medicine, University of PennsylvaniaDepartment of Computer and Information Science, School of Engineering and Applied Science, University of PennsylvaniaDepartment of Computer and Information Science, School of Engineering and Applied Science, University of PennsylvaniaAbstract Despite the success and fast adaptation of deep learning models in biomedical domains, their lack of interpretability remains an issue. Here, we introduce Enhanced Integrated Gradients (EIG), a method to identify significant features associated with a specific prediction task. Using RNA splicing prediction as well as digit classification as case studies, we demonstrate that EIG improves upon the original Integrated Gradients method and produces sets of informative features. We then apply EIG to identify A1CF as a key regulator of liver-specific alternative splicing, supporting this finding with subsequent analysis of relevant A1CF functional (RNA-seq) and binding data (PAR-CLIP).http://link.springer.com/article/10.1186/s13059-020-02055-7Deep learningSplicing codeInterpretationLiver-specific splicingA1CF |
| collection |
DOAJ |
| language |
English |
| format |
Article |
| sources |
DOAJ |
| author |
Anupama Jha Joseph K. Aicher Matthew R. Gazzara Deependra Singh Yoseph Barash |
| spellingShingle |
Anupama Jha Joseph K. Aicher Matthew R. Gazzara Deependra Singh Yoseph Barash Enhanced Integrated Gradients: improving interpretability of deep learning models using splicing codes as a case study Genome Biology Deep learning Splicing code Interpretation Liver-specific splicing A1CF |
| author_facet |
Anupama Jha Joseph K. Aicher Matthew R. Gazzara Deependra Singh Yoseph Barash |
| author_sort |
Anupama Jha |
| title |
Enhanced Integrated Gradients: improving interpretability of deep learning models using splicing codes as a case study |
| title_short |
Enhanced Integrated Gradients: improving interpretability of deep learning models using splicing codes as a case study |
| title_full |
Enhanced Integrated Gradients: improving interpretability of deep learning models using splicing codes as a case study |
| title_fullStr |
Enhanced Integrated Gradients: improving interpretability of deep learning models using splicing codes as a case study |
| title_full_unstemmed |
Enhanced Integrated Gradients: improving interpretability of deep learning models using splicing codes as a case study |
| title_sort |
enhanced integrated gradients: improving interpretability of deep learning models using splicing codes as a case study |
| publisher |
BMC |
| series |
Genome Biology |
| issn |
1474-760X |
| publishDate |
2020-06-01 |
| description |
Abstract Despite the success and fast adaptation of deep learning models in biomedical domains, their lack of interpretability remains an issue. Here, we introduce Enhanced Integrated Gradients (EIG), a method to identify significant features associated with a specific prediction task. Using RNA splicing prediction as well as digit classification as case studies, we demonstrate that EIG improves upon the original Integrated Gradients method and produces sets of informative features. We then apply EIG to identify A1CF as a key regulator of liver-specific alternative splicing, supporting this finding with subsequent analysis of relevant A1CF functional (RNA-seq) and binding data (PAR-CLIP). |
| topic |
Deep learning Splicing code Interpretation Liver-specific splicing A1CF |
| url |
http://link.springer.com/article/10.1186/s13059-020-02055-7 |
| work_keys_str_mv |
AT anupamajha enhancedintegratedgradientsimprovinginterpretabilityofdeeplearningmodelsusingsplicingcodesasacasestudy AT josephkaicher enhancedintegratedgradientsimprovinginterpretabilityofdeeplearningmodelsusingsplicingcodesasacasestudy AT matthewrgazzara enhancedintegratedgradientsimprovinginterpretabilityofdeeplearningmodelsusingsplicingcodesasacasestudy AT deependrasingh enhancedintegratedgradientsimprovinginterpretabilityofdeeplearningmodelsusingsplicingcodesasacasestudy AT yosephbarash enhancedintegratedgradientsimprovinginterpretabilityofdeeplearningmodelsusingsplicingcodesasacasestudy |
| _version_ |
1724472283826749440 |