In reading through the Study section, I came across an area that I thought may be of interest but which was not yet covered by the review-- protein-protein interaction. Since classical machine learning approaches like SVM, random forest, etc. and things like Bayesian networks have been predominant, I think this section would be a potentially meaningful presentation of a good opportunity for deep learning. I found a few papers in which deep learning was applied to this topic and put them below.
- Sequence-based prediction of protein protein interaction using a deep-learning algorithm
- Deep Recurrent Neural Network for Protein Function Prediction from Sequence
- Structure-based prediction of protein–protein interactions on a genome-wide scale
- A novel method based on new adaptive LVQ neural network for predicting protein–protein interactions from protein sequences
One aspect of PPI prediction that stands out is the diversity in features used between different studies. In fact, there was a preprint just released whose purpose is to outline some of the various features for machine learning PPI prediction. This seems like a good indication in my mind that this field would be a prime target for further research that incorporates deep learning.
If this seems like a meaningful subsection to include, perhaps a suitable location would be just below the protein secondary and tertiary structure section. Right here, greenelab/deep-review/sections/04_study.md seems like it would be a logical location.
In reading through the Study section, I came across an area that I thought may be of interest but which was not yet covered by the review-- protein-protein interaction. Since classical machine learning approaches like SVM, random forest, etc. and things like Bayesian networks have been predominant, I think this section would be a potentially meaningful presentation of a good opportunity for deep learning. I found a few papers in which deep learning was applied to this topic and put them below.
One aspect of PPI prediction that stands out is the diversity in features used between different studies. In fact, there was a preprint just released whose purpose is to outline some of the various features for machine learning PPI prediction. This seems like a good indication in my mind that this field would be a prime target for further research that incorporates deep learning.
If this seems like a meaningful subsection to include, perhaps a suitable location would be just below the protein secondary and tertiary structure section. Right here, greenelab/deep-review/sections/04_study.md seems like it would be a logical location.