Benchmark Evaluation of Protein–Protein Interaction Prediction Algorithms

Protein–protein interactions (PPIs) perform various functions and regulate processes throughout cells. Knowledge of the full network of PPIs is vital to biomedical research, but most of the PPIs are still unknown. As it is infeasible to discover all of them experimentally due to technical and resource limitations, computational prediction of PPIs is essential and accurately assessing the performance of algorithms is required before further application or translation. However, many published methods compose their evaluation datasets incorrectly, using a higher proportion of positive class data than occuring naturally, leading to exaggerated performance. We re-implemented various published algorithms and evaluated them on datasets with realistic data compositions and found that their performance is overstated in original publications; with several methods outperformed by our control models built on ‘illogical’ and random number features. We conclude that these methods are influenced by an over-characterization of some proteins in the literature and due to scale-free nature of PPI network and that they fail when tested on all possible protein pairs. Additionally, we found that sequence-only-based algorithms performed worse than those that employ functional and expression features. We present a benchmark evaluation of many published algorithms for PPI prediction. The source code of our implementations and the benchmark datasets created here are made available in open source.

Identification And Expression Analysis of FORMIN Gene Family In Rice (Oryza Sativa)

Background: FORMIN proteins, which are composed of proteins containing FH1 and FH2 domains, play crucial roles in the growth and development of organisms. However, the functions of FORMINs in rice (Oryza sativa) remain largely unclear. Results: In this study, a total of 17 FORMIN genes were identified in rice, OsFH17 was the first time identified in this study. In addition, the distribution on chromosomes, gene structure, as well as conserved motifs of rice FORMINs was investigated. According to their protein structural and phylogenetic features, these 17 rice FORMIN genes were classified into two distinct subfamilies. Subcellular localization prediction showed that rice FORMINs were located in cytosol, golgi, endoplasmic reticulum, extracellular, and vacuole. Protein protein interaction (PPI) prediction results shown that FORMIN protein might answer hormone signals and be involved in cytoskeleton dynamics regulation and cell wall morphology regulation. The results of silico analysis and qRT-PCR confirmation of the gene expression showed that the expression of rice FORMINs were related to their tissue distribution. Moreover, OsFH3, OsFH5 and OsFH7 were upregulated under phytohormone treatments. Conclusions: Overall, our research may shed light on the understanding and further investigation of the biological functions of rice FORMINs.

Network-based protein-protein interaction prediction method maps perturbations of cancer interactome

The perturbations of protein-protein interactions (PPIs) were found to be the main cause of cancer. Previous PPI prediction methods which were trained with non-disease general PPI data were not compatible to map the PPI network in cancer. Therefore, we established a novel cancer specific PPI prediction method dubbed NECARE, which was based on relational graph convolutional network (R-GCN) with knowledge-based features. It achieved the best performance with a Matthews correlation coefficient (MCC) = 0.84±0.03 and an F1 = 91±2% compared with other methods. With NECARE, we mapped the cancer interactome atlas and revealed that the perturbations of PPIs were enriched on 1362 genes, which were named cancer hub genes. Those genes were found to over-represent with mutations occurring at protein-macromolecules binding interfaces. Furthermore, over 56% of cancer treatment-related genes belonged to hub genes and they were significantly related to the prognosis of 32 types of cancers. Finally, by coimmunoprecipitation, we confirmed that the NECARE prediction method was highly reliable with a 90% accuracy. Overall, we provided the novel network-based cancer protein-protein interaction prediction method and mapped the perturbation of cancer interactome. NECARE is available at: https://github.com/JiajunQiu/NECARE.

Large-Scale Protein Interactions Prediction by Multiple Evidence Analysis Associated With an In-Silico Curation Strategy

Predicting the physical or functional associations through protein-protein interactions (PPIs) represents an integral approach for inferring novel protein functions and discovering new drug targets during repositioning analysis. Recent advances in high-throughput data generation and multi-omics techniques have enabled large-scale PPI predictions, thus promoting several computational methods based on different levels of biological evidence. However, integrating multiple results and strategies to optimize, extract interaction features automatically and scale up the entire PPI prediction process is still challenging. Most procedures do not offer an in-silico validation process to evaluate the predicted PPIs. In this context, this paper presents the PredPrIn scientific workflow that enables PPI prediction based on multiple lines of evidence, including the structure, sequence, and functional annotation categories, by combining boosting and stacking machine learning techniques. We also present a pipeline (PPIVPro) for the validation process based on cellular co-localization filtering and a focused search of PPI evidence on scientific publications. Thus, our combined approach provides means to extensive scale training or prediction of new PPIs and a strategy to evaluate the prediction quality. PredPrIn and PPIVPro are publicly available at https://github.com/YasCoMa/predprin and https://github.com/YasCoMa/ppi_validation_process.

Molecular Characterization and Survive Abilities of Salmonella Heidelberg Strains of Poultry Origin in Brazil

The aim of the study was to evaluate the genotypic and phenotypic characteristics of 20 strains of S. Heidelberg (SH) isolated from broilers produced in southern Brazil. The similarity and presence of genetic determinants linked to virulence, antimicrobial resistance, biofilm formation, and in silico-predicted metabolic interactions revealed this serovar as a threat to public health. The presence of the ompC, invA, sodC, avrA, lpfA, and agfA genes was detected in 100% of the strains and the luxS gene in 70% of them. None of the strains carries the blaSHV, mcr-1, qnrA, qnrB, and qnrS genes. All strains showed a multidrug-resistant profile to at least three non-β-lactam drugs, which include colistin, sulfamethoxazole, and tetracycline. Resistance to penicillin, ceftriaxone (90%), meropenem (25%), and cefoxitin (25%) were associated with the presence of blaCTX–M and blaCMY–2 genes. Biofilm formation reached a mature stage at 25 and 37°C, especially with chicken juice (CJ) addition. The sodium hypochlorite 1% was the least efficient in controlling the sessile cells. Genomic analysis of two strains identified more than 100 virulence genes and the presence of resistance to 24 classes of antibiotics correlated to phenotypic tests. Protein-protein interaction (PPI) prediction shows two metabolic pathways correlation with biofilm formation. Virulence, resistance, and biofilm determinants must be constant monitoring in SH, due to the possibility of occurring infections extremely difficult to cure and due risk of the maintenance of the bacterium in production environments.

Predicting Protein–Protein Interactions via Gated Graph Attention Signed Network

Protein–protein interactions (PPIs) play a key role in signal transduction and pharmacogenomics, and hence, accurate PPI prediction is crucial. Graph structures have received increasing attention owing to their outstanding performance in machine learning. In practice, PPIs can be expressed as a signed network (i.e., graph structure), wherein the nodes in the network represent proteins, and edges represent the interactions (positive or negative effects) of protein nodes. PPI predictions can be realized by predicting the links of the signed network; therefore, the use of gated graph attention for signed networks (SN-GGAT) is proposed herein. First, the concept of graph attention network (GAT) is applied to signed networks, in which “attention” represents the weight of neighbor nodes, and GAT updates the node features through the weighted aggregation of neighbor nodes. Then, the gating mechanism is defined and combined with the balance theory to obtain the high-order relations of protein nodes to improve the attention effect, making the attention mechanism follow the principle of “low-order high attention, high-order low attention, different signs opposite”. PPIs are subsequently predicted on the Saccharomyces cerevisiae core dataset and the Human dataset. The test results demonstrate that the proposed method exhibits strong competitiveness.

Prediction of Protein-protein Interactions in Arabidopsis Thaliana using Partial Training Samples in a Machine Learning Framework

Background: Protein-protein interactions (PPI) play a vital role in a wide range of biological processes starting from cell-cell interactions to developmental control in all organisms. However, experimental identification of PPI is often laborious, time-consuming and costly compared to computational prediction. There are several computational prediction models in the literature based on complete training samples, but none of them dealt with the partial training samples. Objective: The objective of this work was to develop an effective PPI prediction model for Arabidopsis Thaliana using partial training samples in a machine learning framework. Methods: We proposed an effective computational PPI prediction model by combining random forest (RF) classifier and autocorrelation (AC) sequence encoding features with 1:2 ratio of positive-PPI and unknown-PPI samples. Results: We observed that the proposed prediction model produces the highest average performance scores of sensitivity (94.62%), AUC (0.92) and pAUC (0.189) with the training datasets and sensitivity (88.14%), AUC (0.89) and pAUC (0.176) with the test datasets of 5-fold cross-validation compared to other candidate predictors based on LDA, LOGI, ADA, NB, KNN & SVM classifiers. It also computed the highest performance scores of TPR (91.82%) and pAUC (0.174) at FPR= 20% with AUC (0.948) compared to other candidate predictors. Conclusion: Overall performance of the developed model revealed that our proposed predictor might be useful to elucidate the biological function of unseen PPIs from a large number of candidate proteins in Arabidopsis thaliana.

“Predictors of pacemaker implantation after TAVI according to kind of prosthesis and risk profile: a systematic review and contemporary meta-analysis”

Background Permanent pacemaker implantation (PPI) may be required after transcatheter aortic valve implantation (TAVI). Evidence on PPI prediction has largely been gathered from high risk patients receiving first generation valve implants. Objectives We undertook a meta-analysis of the existing literature to examine the incidence and predictors of PPI after TAVI according to generation of valve, valve type and surgical risk. Methods We made a systematic literature search for studies with ≥100 patients reporting the incidence and adjusted predictors of PPI after TAVI. Subgroup analyses examined these features according to generation of valve, specific valve type and surgical risk. Results We obtained data from 43 studies, encompassing 29,113 patients. PPI rates ranged from 6.7% - 39.2% in individual studies with a pooled incidence of 19% (95% CI 16-21). Independent predictors for PPI were age (OR 1.05; 95% CI: 1.01-1.09), left bundle branch block (LBBB) (OR: 1.45; 95% CI: 1.12 to 1.77), right bundle branch block (RBBB) (OR: 4.15; 95% CI: 3.23 to 4.88), implantation depth (OR: 1.18; 95% CI: 1.11 to 1.26) and self-expanding valve prosthesis (OR 2.99; 95% CI: 1.39-4.59). Among subgroups analyzed according to valve type, valve generation and surgical risk, independent predictors were RBBB, self-expanding valve type, first degree atrioventricular block and implantation depth. Conclusions The principle independent predictors for PPI following TAVI are age, RBBB, LBBB, self-expanding valve type and valve implantation depth. These characteristics should be taken into account in pre-procedural assessment to reduce PPI rates. PROSPERO ID CRD42020164043.

3D Genome Contributes to Protein-Protein Interactome

Numerous computational methods have been proposed to predict protein-protein interactions, none of which however, considers the original DNA loci of the interacting proteins in the perspective of 3D genome. Here we retrospect the DNA origins of the interacting proteins in the context of 3D genome and discovered that 1) if a gene pair is more proximate in 3D genome, their corresponding proteins are more likely to interact. 2) signal peptide involvement of PPI affects the corresponding gene-gene proximity in 3D genome space. 3) by incorporating 3D genome information, existing PPI prediction methods can be further improved in terms of accuracy. Combining our previous discoveries, we conjecture the existence of 3D genome driven cellular compartmentalization, meaning the co-localization of DNA elements lead to increased probability of the co-localization of RNA elements and protein elements.