Multi-schema computational prediction of the comprehensive SARS-CoV-2 vs. human interactome

Dick, Kevin; Chopra, Anand; Biggar, Kyle K.; Green, James R.

doi:10.7717/peerj.11117

Cited by 10 publications

(13 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…numerical mapping models), but to leverage the context enabled by transfer learning these DTA measures (i.e. CM-RP), promises exciting results given that the application of RP to related bioinformatic problems has led to statistically significant improvements of predictor performance 42 , 43 , 48 .…”

Section: Resultsmentioning

confidence: 99%

“…Prior DTI literature suggests that the incorporation of individual DTI predictors into an ensemble will outperform those individual models 50 . The work of Dick et al on predicting protein-protein interactions (PPI) between SARS-CoV-2 and humans demonstrated that RP could be used to ensemble two PPI predictors (the Protein-protein Interaction Prediction Engine [PIPE4] 51 and the Scoring PRotein INTeractions [SPRINT] 52 models) to produce an RP fusion model 48 .…”

Section: Resultsmentioning

confidence: 99%

“…For similar bioinformatic tasks, the Reciprocal Perspective (RP) framework has been leveraged in a cascade to improve initial model prediction results 42 , 43 . In subsequent work, RP was used for the cascaded combination of multiple experts ( ) 48 , and we here hypothesize that RP can be used not only for the combination of multiple experts ( ) but also as a means of learning a domain transfer model. We depict in Fig.…”

Section: Methodsmentioning

confidence: 99%

“…In previous work, RP was demonstrated to significantly improve the predictive performance of protein-protein interaction predictors 42 and the performance of microRNA target predictors 43 . RP has also been shown capable of fusing component models ( ) as an ensembling method 48 . In this work, we propose to utilize RP as both a transfer learning method (to learn an approximated mapping from to ) as well as a many-CM ensembling method ( ).…”

Section: Methodsmentioning

confidence: 99%

“…These features contextualise each pair among all other predicted pairs thus making use of semi-supervised distributions and a variety of features types (including rank, fold, statistics, and score-types). Thus RP rescores the predicted DTI pair as part of a cascaded super learning model, typically an eXtreme Gradient Boosting (XGBoost) model 49 as in Kyrollos et al and Dick et al 43 , 48 .…”

Section: Methodsmentioning

confidence: 99%

See 4 more Smart Citations

Reciprocal perspective as a super learner improves drug-target interaction prediction (MUSDTI)

Dick

Kyrollos

Cosoreanu

et al. 2022

Sci Rep

Self Cite

View full text Add to dashboard Cite

The identification of novel drug-target interactions (DTI) is critical to drug discovery and drug repurposing to address contemporary medical and public health challenges presented by emergent diseases. Historically, computational methods have framed DTI prediction as a binary classification problem (indicating whether or not a drug physically interacts with a given protein target); however, framing the problem instead as a regression-based prediction of the physiochemical binding affinity is more meaningful. With growing databases of experimentally derived drug-target interactions (e.g. Davis, Binding-DB, and Kiba), deep learning-based DTI predictors can be effectively leveraged to achieve state-of-the-art (SOTA) performance. In this work, we formulated a DTI competition as part of the coursework for a senior undergraduate machine learning course and challenged students to generate component DTI models that might surpass SOTA models and effectively combine these component models as part of a meta-model using the Reciprocal Perspective (RP) multi-view learning framework. Following 6 weeks of concerted effort, 28 student-produced component deep-learning DTI models were leveraged in this work to produce a new SOTA RP-DTI model, denoted the Meta Undergraduate Student DTI (MUSDTI) model. Through a series of experiments we demonstrate that (1) RP can considerably improve SOTA DTI prediction, (2) our new double-cold experimental design is more appropriate for emergent DTI challenges, (3) that our novel MUSDTI meta-model outperforms SOTA models, (4) that RP can improve upon individual models as an ensembling method, and finally, (5) RP can be utilized for low computation transfer learning. This work introduces a number of important revelations for the field of DTI prediction and sequence-based, pairwise prediction in general.

show abstract

Section: Resultsmentioning

confidence: 99%