Inter-Species/Host-Parasite Protein Interaction Predictions Reviewed

Abstract

Background

Host-parasite protein interactions (HPPI) are those interactions occurring between a parasite and its host. Host-parasite protein interaction enhances the understanding of how parasite can infect its host. The interaction plays an important role in initiating infections, although it is not all host-parasite interactions that result in infection. Identifying the protein-protein interactions (PPIs) that allow a parasite to infect its host has a lot do in discovering possible drug targets. Such PPIs, when altered, would prevent the host from being infected by the parasite and in some cases, result in the parasite inability to complete specific stages of its life cycle and invariably lead to the death of such parasite. It therefore becomes important to understand the workings of host-parasite interactions which are the major causes of most infectious diseases.

Objective

Many studies have been conducted in literature to predict HPPI, mostly using computational methods with few experimental methods. Computational method has proved to be faster and more efficient in manipulating and analyzing real life data. This study looks at various computational methods used in literature for host-parasite/inter-species protein-protein interaction predictions with the hope of getting a better insight into computational methods used and identify whether machine learning approaches have been extensively used for the same purpose.

Methods

The various methods involved in host-parasite protein interactions were reviewed with their individual strengths. Tabulations of studies that carried out host-parasite/inter-species protein interaction predictions were performed, analyzing their predictive methods, filters used, potential protein-protein interactions discovered in those studies and various validation measurements used as the case may be. The commonly used measurement indexes for such studies were highlighted displaying the various formulas. Finally, future prospects of studies specific to human-plasmodium falciparum PPI predictions were proposed.

Result

We discovered that quite a few studies reviewed implemented machine learning approach for HPPI predictions when compared with methods such as sequence homology search and protein structure and domain-motif. The key challenge well noted in HPPI predictions is getting relevant information.

Conclusion

This review presents useful knowledge and future directions on the subject matter.

1. INTRODUCTION

Recently, infectious diseases became a major health concern to the world as a result of several millions of sicknesses and deaths recorded every year, despite clinical efforts and development. Several research works tried to identify host-parasite interactions from different perspectives to get a thorough understanding of parasite and defensive means to curb parasitic infections. In host-parasite interactions, proteins are involved, and these proteins control all biological systems in a cell such as molecular functions and biological processes. Proteins interrelate with other proteins to generate a protein interaction network. Therefore, interactions between proteins become vital keys in various biological activities.

Host-parasite protein interactions play critical roles in host and parasite infection and therefore identifying the protein-protein interactions (PPIs) that allow parasites infect the host would be of great assistance in discovering potential drug targets [1]. Unfortunately, there is a limitation in the presently available knowledge on genes and proteins implicated in such interactions because only a few works have been experimentally proven as protein interactions between host and pathogen in several host-parasite systems. Computational method of identifying host-parasite protein interactions became popular because of this limitation in experimental method. The computational approaches are divided basically into methods based on; sequence homology, protein structure, domain and motif and machine learning [2].

2. Methods of host-parasite protein interactioN predictions used in literature

There are several computational approaches used for predictions, but the popularly used ones in inter-species/host-parasite protein interaction predictions are presented here. They are sequence homology search, domain-motif, structure-based and machine learning approaches. Fig. (1) presents a graphical view of such methods. Fig. (2) is a graphical view of features exploited by previous studies in host-parasite/inter-species protein-protein interaction predict-tions.

Fig. (1).

Methods used in Literature to predict Host-parasite protein interactions.

Fig. (2).

Exploited features for HPPI and Inter-species prediction from literature.

2.1. Sequence Homology-Based Method

The homology-based method is a traditional means of intra-specie protein-protein interactions prediction. This method has been embraced in the prediction of inter-species host-parasite protein-protein interactions. The idea behind sequence homology method is that the interaction between two proteins in a particular species will probably be conserved in related species [3]. The conserved interactions are called Interologs. What this means is that pairs of homologous protein originate from similar family pairs of interacting proteins which will likely take the structure and function as well as interactions of the family proteins. The operation of homology method for identification of inter-species protein-protein interaction involves first getting the template PPI pair (x y), and finding the homolog x' in the host and the homolog y' in the parasite and then infer that (x', y') interact [3]. Some of studies utilized this method as a method of host-parasite protein interactions prediction. Such studies include; [4-13]. The limitation of this method of protein interaction prediction is that inferences cannot be made about interactions between specie-specific families of genes.

2.2. Domain and Motif Interaction-Based Method

Domains are the core determinants of the structure and function of proteins which perform specific function in facilitating proteins interaction with other molecules [14]. Majority of PPIs are facilitated by domain-motif interaction by binding domains in protein to short linear motifs in interacting partners. These interactions are mostly implicated in major cellular processes, requiring their tight regulation [15]. A number of studies have used this method as a building block for predicting protein-protein interactions in single species [16-18]. Study [1] happens to be the first to explore this method for inter-species protein interaction prediction, although, the number of interactions predicted was few and their biological significance has not been assessed. A procedure that predicts interacting protein between parasite and host from the integration of protein domain profiles and interacting protein between proteins from the same organism was used by [1].

Here, Bayesian statistics were used to find the probability of proteins interactions for every pair of functional domains found in the protein pair. This procedure was applied in the host-parasite system by identifying domains in the individual host and parasite protein having at least one domain. Then, the probability of the interaction was computed for each pair of host and parasite with at least one domain. Other studies [19-22] also used domain and motif interaction based method for HPPI.

2.3. Structure-Based and Structural Modeling Approaches

In the structure-based method of prediction, two proteins with similar structures to a known proteins interaction pairs are likely interacting in a structurally similar way. Quite a number of studies already employed the structural information to predict the similarities between query proteins such as host-parasite protein interaction and template PPIs to infer that interaction exists in host-parasite protein pairs that match some template PPIs [23]. Doolittle and colleague [24] constructed a map interaction of HIV-1 and human. The same Doolittle and colleague in another study [25] applied similar approach for building a network of interactions to Dengue virus and its host. A study by [26], presumes interactions between proteins that are structurally homologous using Human and Influenza A NS1. Protein structure-based protocols were used by [27] to discover potential protein interactions in P. falciparum and host erythrocytes. For structural modeling, study [28] conducted HPPI comparative modeling of 3D structures and applied the techniques to 10 pathogens including Mycobacterium, apicomplexa, kinetoplastida and Plasmodium falciparum which are responsible for neglected human diseases.

2.4. Machine Learning

Machine learning approach is a robust method of HPPI predictions, although this method has not been used extensively in HPPI prediction.

Random Forest (RF) is a classifier algorithm made up of decision trees. Individual tree in the training phase is built by random feature vector sampled from a dataset independently. A little part of the variables is selected at random and individual classification tree is raised for every node in a tree. In order to group a new object, the input vector is set up for each of the trees in the forest. Based on the largest vote, a class is allocated to the object. Random forest is a practical classifier when large datasets with a large number of features are concerned because there will be no need to feature select or feature delete. The Random forest can also classify features based on importance and can also be used to recover missing data. Nevertheless, specific databases with noisy data may overfit. Random forest and Decision trees are extensively employed in bioinformatics and computational biology for classifying biological data [29] especially for PPI prediction [30]. The RF approach was used by study [31] for feature evaluation in order to accurately predict protein-protein interactions from negative dataset. In predicting cytokine-receptor interactions, Random forest classifier method was used by [32]. The studies that implemented Random forest classifier approach to predict host-parasite/inter-species PPIs that this work is interested in are studies [19, 33]. Study [19] employed RF method for HPPI predictions by integrating thirty-five features within eight groups. The same random forest classifier was used by [33] to assess the quality of conserved interactions for predictions of putative PPIs between human host and malaria parasite.

Support vector machine (SVM) is another classifier approach used in HPPI predictions. SVM classifier is reinforced by margin optimization. This margin for an object is connected to the certainty of its classification. Objects with correct classification will have large margins and objects with classification unclear will probably have small margins. Training SVM can be achieved using a labeled dataset with each data labeled to indicate that it belongs to a class or two. Support vector machine is a very powerful classification method for arbitrary complex problems. Studies [34-38] employed this approach. Here, fixed feature vector length that connotes relative frequency of conserved amino acids in the protein sequence was used.

The other machine learning approaches employed in Literature include; Expectation maximization algorithm by [37]. Here, the study predicted, HPPI in human RBCs and merozoite membrane proteins from estimates of domain-domain interaction probabilities. Also, study [38], proposed Multi-instance transfer learning method called AdaBoost which was used to re-construct the proteome-wide Salmonella and human protein-protein interaction networks. The training dataset was improved using homolog knowledge transfer in the form of independent homolog instances. AdaBoost instance re-weighting was employed to offset the noise from homolog instances. Fig. 1 is a graphical display of computational methods used in literature for such predictions.

2.5. Strength of Methods Reviewed

Homology-based method: The simplicity and the seeming biological basis of homology-based method is a major advantage. Prediction using this method only requires template PPIs and protein sequences data. The method is also scalable with application to several host-parasite systems. Studies [4, 6, 7], used homology-based approach alone for prediction while [33] combined homology-based method with other methods to predict host-parasite protein-protein interaction.

Domain-domain/motif: Protein domain is a key in protein structure prediction. In order to also determine protein structure, annotate functions, mutagenesis analysis, and protein engineering, protein domain prediction is important. Ability to predict domains from sequence information increases identification of tertiary structure; improve annotation of protein function, assist the determination of structure and give direction to engineering and mutagenesis of protein. The identification of domains within a protein sequence also serves as a foundation for other methods.

Structure-based prediction approach: This method of prediction is quite efficient when the structure of the target protein has not been resolved experimentally. The similarity between structure leads to an identification of homologs.

Machine learning is a group of computational methods used to identify complicated patterns in a given dataset and make decisions on previously unseen data.

3. tabulations of hppi predictions in human host & plasmodium falciparum and other inter-species protein interaction predictions

This part is divided into two sections. Section 3.1 presents specific studies conducted on human host and Plasmodium falciparum parasite PPI predictions. It looks at the method of prediction by each study, the predictive method, filters used, potential PPIs identified and measurements used. Table 1 is a tabular presentation of section 3.1.

Table 1.

Cross section of studies on human host and Plasmodium falciparum PPIs predictions

Host-Parasite	Predictive Method	Potential PPIs Identified	Filter	Measurements	Ref
Human-Plasmodium falciparum	Combined interaction probability of domains Bayesian statistics for assessment	A total of 516 PPIs between human and Pf were predicted. Important PPIs predicted are PfEMP1s and MSP1s, Q8IAS3, plasminogen (Q5TEH4) and pfEMP1, Q8IAL6 and Q8I339. They all interact with human blood coagulation proteins which may play a role in disrupting human blood coagulation pathways. Q8IHZ5, a known subtilisin-like protease, interacts with a number of blood coagulation proteins, which suggests that it may be involved in the degradation of blood platelets. Also, hypothetical Plasmodium protein Q8IKP8 interacts with the predicted partners of Q8IHZ5.	Gene ontology terms	Area under the Curve (AUC), Sensitivity and Precision	[1]
Human-Plasmodium falciparum	Interlogs inferred from ortholog information	Interactions between putative HSP40 homologs of P. falciparum and the H. sapiens TNF receptor associated factor family was revealed here, suggesting a role for these interactions in the interference of the human immune response to P. falciparum. Calmodulin (PF14_0323), interacts with 50 human proteins. Among the 50 human proteins interacting with PF14_0323, thirteen (13) of them interact with human calmodulin (CALM3). This suggests that P. falciparum calmoduin shares some of the targets of human calmodulin, and may hijack these PPIs for its purpose. PF14_0359 and the TNF receptor associated factor family (TRAF1, TRAF2 and TRAF6) are predicted also to interact.	Gene ontology annotations and Presence/ absence of translocational signals.	Sensitivity	[4]
Human-Plasmodium falciparum	Homology detection method using template PPI databases, DIP, and iPfam	Remarkable interactions are: Plasmepsins and host cytoskeletal proteins, interaction between TRAP and ICAMs	Database of Interacting Proteins (DIP) sequences		[5]
Human-Plasmodium falciparum	Interolog	The study observed that most of the highly interacting proteins were involved in structural assembly of the pathogen such as actin, tubulin, and histone. 𝛼-tubulin was finalized as an important protein involved in the infection process.	Cellular location, Gene ontology, and Functional role.		[13]
Human-Plasmodium falciparum	Homology-based approach	A total of 208 physicochemically viable interactions were predicted. The key interacting proteins are: SAR1 and the host ADP-ribosylation factor-binding protein GGA3 (Q9NZ52), Host calcium-activated potassium channel protein 4, KCNN4 (UniProt ID: O15554), and conserved parasitic protein of unknown function, PF3D7_1463900.	Intrachain heterodomain interactions from iPfam, Intra host and intra pathogen interactions and Expression profile of parasite proteins from PlasmoDB.		[27]
Human-Plasmodium falciparum	Comparative Modelling	The key prediction from this study relating to P. falciparum and human are; P. falciparum thrombospondin-related adhesive protein (TRAP, SSP2, PF13_0201) interact with human Toll-like receptor 4 (TLR4, ENSP00000346893), based on a template structure of Glycoprotein IBa bound to Von Willenbrand factor (PDB1M10). TRAP, animmunogenic protein used as a component of several vaccine candidates, interacts with TLR4.	Biological context and Network-level Information		[28]
Host-Parasite	Predictive Method	Potential PPIs Identified	Filter	Measurements	Ref
Human-Plasmodium falciparum	Sequence Orthology/ Homology and Random forest	The discovery here is that parasite proteins predominantly target central proteins to take control of a human host cell. Several prominent pathways of signaling and regulation proteins were predicted to interact with parasite chaperones.	Expression data and molecular properties	Area under the curve (AUC)	[33]
Human-Plasmodium falciparum	Estimation maximization	A network consisting of 205 PPIs between parasite and human membrane proteins were predicted. A further prediction shows that SNARE proteins of parasites and APP of humans may function in the invasion of RBCs by parasites.	Gene expression data	Area under the curve (AUC)	[37]
Human-Plasmodium falciparum	Mining of combined HPPI data	The analysis in the study revealed; apolipoproteins and temperature/Hsp expression on PfEMP1 presentation, the essence of MSP-1 in platelet activation, role of parasite proteins in TGF-β regulation and the contribution of albumin in astrocyte dysfunction.	Gene Ontology, Tissue-specific annotation		[39]

The second section, 3.2 reviews other studies on host/parasite PPI predictions and inter-species PPIs, presenting the type of host, pathogen or species used in the study. The methods of prediction employed the predictive method, filters used, potential PPIs identified and measurements used as in section 1. Both sections are presented in the form of a table. Table 2 is a tabular representation of section 3.2.

Table 2.

Cross section of other studies on host-parasite/inter-species protein-protein interaction predictions.

Host- Pathogen	Predictive Method	Potential PPIs Identified	Filter	Measurements		Ref.
Mycobacterium Tuberculosis – Homo Sapien	Homology detection approach based on sequence motif	A total of 118 pairs of HPIs were obtained from 43 Mycobacterium tuberculosis proteins and 48 Homo sapiens proteins were predicted and stored in the PATH database	Domain-Domain Interactions (DDIs), and Functional annotations of protein and publicly available experimental results for further filter		F1 Score	[7]
Salmonella-Human	Sequence and interacting domain similarity approach	This study predicted 29 out of 59 gold standard PPIs used. With Domain-based prediction feature, nine (9) of the gold standard interactions were predicted. These nine interactions are also part of the set of 29 PPIs formerly predicted.	Domain-based prediction feature			[8]
Candida albicans-Zebrafish	Ortholog-based PPIs and Multivariate linear dynamic model of regulatory responses	This study developed a computational framework. Some of the predictions between Candida albicans-Zebrafish were done. An important discovery here is that redox status is critical during the battle between the host and pathogen, which could determine the outcome of infection.	Sequence-targeted probes derived from the individual genome			[11]
Mycobacterium Tuberculosis H37Rv –Human	Stringent homology-based approach	An interesting discovery made aside from PPI predictions include host proteins and pathogen proteins that partake in the host-pathogen PPIs which tend to be hubs in their own intra-species PPI network. Again, host and pathogen proteins that are involved in host-pathogen PPIs might have a lengthier primary sequence, more domains, more hydrophilic and others.	PATRIC database			[12]
HIV 1- Human protein	Supervised learning using Random Forest Classifier	A key prediction from this study is HIV-1 protein tat and human vitamin D receptor (VDR) Tat is a regulatory protein of HIV-1. The interaction has also been validated experimentally.	Eukaryotic Linear Motif (ELM) database		ROC-AUC, Precision-Recall	[19]
Host- Pathogen	Predictive Method	Potential PPIs Identified	Filter	Measurements		Ref.
Human- microbial Oral	Ensemble methodology for prediction naïve Bayes classifier for training and validation	The study revealed important pathways involved in the onset of infectious oral diseases, and also potential drug-targets and biomarkers. Also, the first computational model of the Human-Microbial oral interactome was constructed.	PPI pairs from the five databases		Area under the ROC-AUC, F1, score, Accuracy, Precision-Recall	[22]
HIV virus-Human	Structural similarity	A total of 502 interactions involving 137 human proteins were predicted. Three interactions consistent with two other studies predicted by this study are; gp41 and LCK, gp41 and PLK1, IN and XPO1. Twenty-two (22) of true positives predictions out of 265 predictions were made.	RNAi functional data and shared Gene Ontology cellular component annotation for further filter			[24]
Dengue virus-Human and Insect hosts	Structural similarity	They predicted 2,073 interactions among viral and human proteins and found 7 out of 19 true positives. The study revealed the possibility that some of the protein interactions which enable DENV to manipulate the cellular pathways of two hosts are conserved between the species.	Functional information from recent literature and Gene Ontology cellular component annotation (Subcellular co-localiza tion)			[25]
Influenza A NS1–Human	Method based on structural homologous proteins interactions	The study predicted that out of 41 human proteins of influenza–human PIN, twelve (12) have been identified to be host factors for influenza virus replication. When Influenza–human interactome were combined with predicted and literature data, forty-seven (47) of 364 human proteins were identified to be host factors directly controlling viral replication	Predicted and literature data			[26]
Human-Human papillomaviruses (HPV) and hepatitis C virus (HCV)	Support vector machine	This study predicted interactions between viruses and human proteins. The comparative analysis of HCV and HPV viral interaction networks gave 11 common human proteins that are targeted by both viruses. The SVM model revealed an average accuracy of 81.6% to predict human-HCV proteins, and accuracy of 83.3% to predict Human-HPV proteins.	BLAST and Gene Ontology		Sensitivity, specificity accuracy	[34]
Human- Yersinia Pestis, Francisella Tularensis, Salmonella and Bacillus anthracis	Multitask learning approach	The study carried out a host-pathogen protein-protein interaction (PPI) prediction involving a fixed host and pathogens with various bacterial species. A set of interologs were predicted to exist between the four datasets.	BLAST		Precision-Recall and F1 score	[35]
HIV-1 and Human	Ensemble Transfer Learning method and Support Vector Machine for classification	The study deployed a model that is robust against data unavailability with less demanding data constraint. Analysis of overlapped predictions between the model in this study and the other existing models were carried out and the model was applied to novel host- pathogen PPIs identification.	Gene Ontology		ROC-AUC, F1 Score, Precision-Recall	[36]
Host- Pathogen	Predictive Method	Potential PPIs Identified	Filter	Measurements		Ref.
Human T-cell leukemia viruses (HTLV) retroviruses-Human	Multi-instance Ada boost transfer learning method	The study used homology knowledge (GO) in the form of auxiliary homolog instance to address the problem of scarcity and unavailability of data. The study concluded that the method is effective in enriching information abundance evidenced by the HTLV-human PPI networks predicted.	AdaBoost instance reweighting		ROC-AUC, Precision recall Curve, Specificity, Sensitivity and F1 score	[38]
Xanthomonas oryzae pathovar oryzae (Xoo) oryzae-Rice	XooNET uses Structural Interactome MAP (PSIMAP), Protein interactions Experimental Interactome MAP (PEIMAP) and Domain-Domain interactions from iPfam	This study discovered 15 annotated AvrBs3 homologues in Xoo; Xoo1125, a hypothetical protein has over 60 interaction partners including the Avr proteins, responsible for the loss of pathogenicity when transposon insertion eukaryotic linear motifs	Psi-Blast and hmm pfam for domain assignment			[40]
HIV1-Human	Method of domain-motif based on Multiple sequence alignments	The study predicted 109 true positives HPPIs from a total of 4,523 predictions. A total of 56 of the 133 Eukaryotic Linear Motif (ELM) resource were conserved on some HIV-1 protein. The essential discovery here is that ELMs that are conserved may appear frequently on human proteins. ELM LIG_PDZ_3 occurred on 90% of human proteins while some other ELMs such as LIG_EH1_1, occurred on quite a few human proteins.	Conserved Eukaryotic Linear Motifs (ELMs) in Protein's multiple alignments			[41]
Dengue virus-Human	Domain and motif based method	A total of 79 human proteins (out of 1654) were identified to have interactions with viral proteins deposited in the VirHostNet database. The Functional enrichment analysis of the remaining 1,574 human proteins revealed 1,224 proteins that share biological processes annotations with the 79 identified human proteins targeted by the virus.	Human domain set was used to filter the 3DID database in order to obtain motif-domain interactions involving only domains in the human proteome			[42]
Plasmodium berghei-Mouse	Correlated gene expression profiles	The first network of mouse/mosquito malaria host-parasite interactions was predicted in the study. Several host genes involved in malaria infections were discovered. Specific ones include chromatin remodeling which is important for malaria interacting with its host to control gene expression timing. Also, genes involved in vesicle transport to the Golgi are important in host–parasite interactions for both Plasmodium and mouse especially to export proteins to the host cell surface.	Yeast two-hybrid interologues and CSS interactions			[43]
Hepatitis C virus (HCV)-Human	Method based on Domain-domain interactome. Topological and functional analysis of the network.	Domain-centric perspective was used to construct a global landscape of virus-host interactions. The study identified that viruses use unique domains to interact the same host partners with fundamental functions and it also employed conserved DDIs occurring in host interactomes to mediate the interspecies interaction.	Integrated domain-domain interaction (IDDI) database			[44]
Host- Pathogen	Predictive Method	Potential PPIs Identified	Filter	Measurements		Ref.
Mycobacterium tuberculosis - Homo sapiens	Interolog method and domain-domain interactions to filter HPPIs	The study predicted 118 pairs of HPIs. A biological interaction network between M. tuberculosis and Homo sapiens was then constructed using the predicted inter- and intra-species interactions based on the 118 pairs of HPIs. Finally, a web accessible database named PATH was built.	Protein sequences and Functional annotations of protein and publicly available experimental results			[45]
Francisella-human	Comparative genomics and Literature	The study identified 222 unique PPIs between 18 Francisella tularensis proteins and 183 human proteins. Twelve (12) Human-F. tularensis interactions were chosen for re-testing. There was a confirmation of four interactions in this assay. They are FTT0482c with WD repeat-containing protein 48 (WDR48), FTT1538c with 78-kDa glucose-regulated protein (HSPA5), FTT1538c with WDR48, and FTT1597 with AP-3 complex subunit mu-1 (AP3M1).	Proteome-scale yeast two-hybrid (Y2H)			[46]
Human-Staphylococcus aureus and Human-Neisseria. Meningitides	Genome-wide protein microarray analysis	The study revealed interactions between the S. aureus immune evasion protein FLIPr (formyl-peptide receptor like-1 inhibitory protein) and the human complement component C1q, as key players of the offense-defense fighting; and of the interaction between meningococcal NadA and human LOX-1 (low-density oxidized lipoprotein receptor), an endothelial receptor	Human recombinant proteins from the GNF library			[47]
Grass carp-grass carp reovirus (GCRV)	Structural motif-domain interactions	A systems-based framework for the understanding of the GCRV infectome and diseasome was provided by the study. JAM-A protein was predicted to interact with GCRV Sigma1-like protein motifs, sharing similar binding mode compared with orthoreovirus.	RNA-seq data from previous work			[48]
Human- human immunodeficiency virus 1 (HIV-1).	Short linear motifs	A method that predicts virus-host SLiM mediated PPIs and rank ranks candidate interactions were developed. The study discovered that the majority of conserved linear motifs in the HIV-1 virus are located in disordered regions.	NIAID HIV-1-human interactions and the set of ELM mediated HIV-1-human interactions.			[49]
Human- Mycobacterium tuberculosis	Pairwise structure similarities	Secreted proteins of the STPK, ESX-1, and PE/PPE family in M. tuberculosis targeted human proteins involved in immune response and phagocytosis. M. tuberculosis also targeted host factors known to regulate HIV replication	Cellular localization information			[50]

3.1. Tabulation of Studies on Human Host and Plasmodium falciparum PPI Predictions

Table 1 is a tabular cross section of results from literature of host-parasite protein-protein interaction predictions specifically for human host and Plasmodium falciparum parasite. The table highlights the methods used and also described the work done.

3.2. Tabulation of Studies on Other Host-Parasite/Inter-Species Protein Interaction Predictions

Table 2 presents other studies on inter-species/host-parasite protein interaction predictions in literature, the methods of prediction employed, the filters used, potential Protein-protein Interactions identified and measurements used.

4. some MEASUREMENTS INDEXES used in hppi predictions and future prospects

4.1. Some Evaluation Measurements Indexes used in HPPI Predictions

Common measurements indexes identified in literature for host-parasite protein-protein interactions include; Sensitivity as used by [1, 4, 34] and Specificity by [34], Accuracy by [22, 34], AUC (area under the roc curve) by [1, 22, 33, 36-38]; Precision-Recall by [1, 22, 36, 37] and F1score by [7, 22, 34].

The formula for calculating the evaluation indexes for Sensitivity, Specificity, Accuracy, ROC-AUC and Precision-Recall are;

$$\mathit{Sensitivity} [\mathit{Recall}]\; =\frac{\mathit{TP}}{\mathit{TP}+\mathit{TN}}$$ (1)

$$\mathit{Specificity} =\frac{\mathit{TN}}{\mathit{TN}+\mathit{FP}}$$ (2)

$$\mathit{Accuracy} =\frac{\mathit{TP}+\mathit{TN}}{\mathit{TP}+\mathit{FP}+\mathit{TN}+\mathit{FN}}$$ (3)

$$\mathit{Precision} =\frac{\mathit{TP}}{\mathit{TP}+\mathit{FP}}$$ (4)

$$\mathit{F1} \mathit{Score} =\frac{2x\mathit{Precision}x\mathit{Recall}}{\mathit{Precision}+\mathit{Recall}}$$ (5)

$$\mathit{AUC} = \mathit{Area} \mathit{under} \mathit{ROC} \mathit{Curve}$$ (6)

4.2. Future Prospects

From the above review, it is eminent that HPPI predictions will uncover the required knowledge in gaining understanding and discovering the crucial interactions that might be responsible for diseases. Such computational studies will shed more light on taking experimental work further and also assist in therapeutic development. It will further enhance deeper insight into the disease under study and open rooms of opportunities to explore new therapy development through the important protein interactions and proteins predicted. The need to explore machine learning methods is proposed here to ensure efficient results.

The future prospects in studies carried out on HPPI for human, and Plasmodium falciparum include; the need to include reliability assessment of protein-protein interactions identified through high-throughput screens in study [1]. For study [4], multiple methods could be combined to improve sensitivity and identify drug targets for drug discovery and design. The extent to which the method by [5] can predict HPPI should be investigated further, through a comprehensive experimental database for PPIs between P. falciparum and human. There is the need to identify molecular strategies that allow the parasite, Plasmodium falciparum to have power over the host in study [33]. This will expand the knowledge of the parasite unique re-modelling processes of the host cell and give potential leads to disease mediation.

Furthermore, in study [36], more studies to understand the significant relationship between the SNPs and parasite invasion are required. There is also the need to experimentally confirm the identified interactions. For study [27], experimental attempts are required to understand possible means of pathogenesis of red blood cells protein-protein interactions predicted in the study. Finally, an in vivo testing and validation of inhibitors identified in study [13] is required to understand the properties as antimalarial drugs

CONCLUSION

Host-parasite interaction prediction is gaining more ground in the recent years because the knowledge from it can give a better understanding of how parasite infects its host, which protein interactions are important for such infection to take place and also identifying drug targets. Computational methods will therefore, play an essential part in creating pathways for experimental HPPI validation by identifying and presenting important interactions that could then be taken further experimentally. In this review, we have been able to identify up to date studies conducted in host-parasite protein interactions with a table showing the host-pathogen involved in the interaction, the methods such studies used in conducting their studies, filters employed and measurements as the case may be. The future prospects of the studies were also mentioned with a specific focus on human-Plasmodium falciparum predictions. This kind of review will help researchers who want to work in this area to know what has been done so far and thus give direction to further studies.

Consent for Publication

Not applicable.

Conflict of Interest

The authors declare no conflict of interest, financial or otherwise.