4 research outputs found
Sequence-based prediction of protein protein interaction using a deep-learning algorithm
No full textBackground: Protein-protein interactions (PPIs) are critical for many biological processes. It is therefore important to develop accurate high-throughput methods for identifying PPI to better understand protein function, disease occurrence, and therapy design. Though various computational methods for predicting PPI have been developed, their robustness for prediction with external datasets is unknown. Deep-learning algorithms have achieved successful results in diverse areas, but their effectiveness for PPI prediction has not been tested. Results: We used a stacked autoencoder, a type of deep-learning algorithm, to study the sequence-based PPI prediction. The best model achieved an average accuracy of 97.19% with 10-fold cross-validation. The prediction accuracies for various external datasets ranged from 87.99% to 99.21%, which are superior to those achieved with previous methods. Conclusions: To our knowledge, this research is the first to apply a deep-learning algorithm to sequence-based PPI prediction, and the results demonstrate its potential in this field.National Natural Science Foundation of China [21673010, 81273436]; Ministry of Science and Technology of China [2016YFA0502303]SCI(E)ARTICLE1
Dimension conversion and scaling of disordered protein chains
No full textTo extract protein dimension and energetics information from single-molecule fluorescence resonance energy transfer spectroscopy (smFRET) data, it is essential to establish the relationship between the distributions of the radius of gyration (R-g) and the end-to-end (donor-to-acceptor) distance (R-ee). Here, we performed a coarse-grained molecular dynamics simulation to obtain a conformational ensemble of denatured proteins and intrinsically disordered proteins. For any disordered chain with fixed length, there is an excellent linear correlation between the average values of R-g and R-ee under various solvent conditions, but the relationship deviates from the prediction of a Gaussian chain. A modified conversion formula was proposed to analyze smFRET data. The formula reduces the discrepancy between the results obtained from FRET and small-angle X-ray scattering (SAXS). The scaling law in a coil-globule transition process was examined where a significant finite-size effect was revealed, i.e., the scaling exponent may exceed the theoretical critical boundary [1/3, 3/5] and the prefactor changes notably during the transition. The Sanchez chain model was also tested and it was shown that the mean-field approximation works well for expanded chains.Ministry of Science and Technology of China [2015CB910300]SCI(E)[email protected]
Additional file 1 of Prediction of liquid–liquid phase separating proteins using machine learning
No full textAdditional file 1: Fig. S1. Snapshot of main page of PSPredictor web server; Fig. S2. PCA 2D projection of PSPs and non-PSPs with three different sets of negative samples; Table S3. Enrichment GO terms of human PSPs predicted by PSPredictor; Table S4. GO term clusters of PSPs with similar meaning in biology
Additional file 2 of Prediction of liquid–liquid phase separating proteins using machine learning
No full textAdditional file 2: Table S1. All models’ training results; Table S2. The training results of three repeats of models with (1) w2v coded, (2) the ratio of positive samples and negative samples is 1:1, (3) sequence number is 586, and, (4) GBDT trained
