Deep-Learning-Based Drug–Target Interaction Prediction

[enricoferrero]

http://doi.org/10.1021/acs.jproteome.6b00618 (http://pubs.acs.org/doi/abs/10.1021/acs.jproteome.6b00618)

Identifying interactions between known drugs and targets is a major challenge in drug repositioning. In silico prediction of drug–target interaction (DTI) can speed up the expensive and time-consuming experimental work by providing the most potent DTIs. In silico prediction of DTI can also provide insights about the potential drug–drug interaction and promote the exploration of drug side effects. Traditionally, the performance of DTI prediction depends heavily on the descriptors used to represent the drugs and the target proteins. In this paper, to accurately predict new DTIs between approved drugs and targets without separating the targets into different classes, we developed a deep-learning-based algorithmic framework named DeepDTIs. It first abstracts representations from raw input descriptors using unsupervised pretraining and then applies known label pairs of interaction to build a classification model. Compared with other methods, it is found that DeepDTIs reaches or outperforms other state-of-the-art methods. The DeepDTIs can be further used to predict whether a new drug targets to some existing targets or whether a new target interacts with some existing drugs.

This could be added in the treat section, subsection patient treatment. Related to this, have you considered whether drug repurposing should have its own subsection?

[agitter]

Thanks, I agree this fits in Treat. I thought we had a placeholder in the outline for drug repurposing, but apparently we don't. It would be a good category to include if someone (you?) is able to draft it before our cutoff date for new sections and we think that deep learning has made important contributions relative to other machine learning approaches.

I've been logging some drug-target interaction papers as issues but haven't written anything about them. I've personally been most excited about new feature representations for chemicals for virtual screening (#313) and focused there first. My intention was to quickly refer to a couple drug-target interaction papers, but that could be a larger section if someone wants to draft it.

[enricoferrero]

OK, I'm on it, there are a couple of nice papers in the drug repositioning area. Hope to have something ready by Monday!

[agitter]

Great! Can you stop by #188 and add a quick note that you intend to write something on the subject? I've been pointing people there for a current status on what has been written, is in progress, and needs to be written.

[enricoferrero]

The following info might be useful if we want to draw parallels between drug-target interaction methods and ligand-based activity screening in terms of observations and features: DeepDTIs uses extended connectivity fingerprints (ECFPs) as features for its 1412 compounds and protein sequence composition (PSCs) features for its targets (1520).