Abstract

   The study of molecular host–parasite interactions is essential to
   understand parasitic infection and adaptation within the host system.
   As well, prevention and treatment of infectious diseases require a
   clear understanding of the molecular crosstalk between parasites and
   their hosts. Yet, large-scale experimental identification of
   host–parasite molecular interactions remains challenging, and the use
   of computational predictions becomes then necessary. Here, we propose a
   computational integrative approach to predict host—parasite
   protein—protein interaction (PPI) networks resulting from the human
   infection by 15 different eukaryotic parasites. We used an
   orthology-based approach to transfer high-confidence intraspecies
   interactions obtained from the STRING database to the corresponding
   interspecies homolog protein pairs in the host–parasite system. Our
   approach uses either the parasites predicted secretome and membrane
   proteins, or only the secretome, depending on whether they are uni- or
   multi-cellular, respectively, to reduce the number of false
   predictions. Moreover, the host proteome is filtered for proteins
   expressed in selected cellular localizations and tissues supporting the
   parasite growth. We evaluated the inferred interactions by analyzing
   the enriched biological processes and pathways in the predicted
   networks and their association with known parasitic invasion and
   evasion mechanisms. The resulting PPI networks were compared across
   parasites to identify common mechanisms that may define a global
   pathogenic hallmark. We also provided a study case focusing on a closer
   examination of the human–S. mansoni predicted interactome, detecting
   central proteins that have relevant roles in the human–S. mansoni
   network, and identifying tissue-specific interactions with key roles in
   the life cycle of the parasite. The predicted PPI networks can be
   visualized and downloaded at [29]http://orthohpi.jensenlab.org.

   Keywords: computational biology, systems biology, biological networks,
   parasitology, schistosomiasis, host–parasite interactions

Introduction

   Parasites are responsible for many diseases that result in millions of
   deaths each year. For instance, the World Health Organization published
   data in 2016 estimating that Plasmodium falciparum alone was
   responsible for around 214 million malaria cases, and 438,000 deaths
   worldwide ([30]1). As well, around 7 million people worldwide were
   reported to be infected with Trypanosoma cruzi, which causes Chagas
   disease that results in life-long morbidity and disability and more
   than 7,000 deaths per year ([31]1). Another highly prevalent disease,
   Leishmaniasis accounts for 20 to 30 thousand deaths a year and is
   caused by protozoan parasites of the Leishmania genus ([32]1).
   Similarly, Schistosomiasis, a neglected parasitic disease of high
   relevance in this work, is mainly caused by five species of the genus
   Schistosoma. The disease has an estimated prevalence of 200 million
   cases worldwide ([33]2). The available treatment for schistosomiasis is
   limited, and the development of resistance is a concern. Thus, there is
   an urgent need to develop novel drugs or vaccines.

   The development of vaccines or treatments has been impeded by the lack
   of understanding of the parasites infection and survival mechanisms.
   Typically, parasites have complex life cycles with several
   morphological stages and infect distinct host cell types and tissues.
   For that, parasites display a resourceful capacity to live in different
   environmental conditions (intra and extracellular parasites) and also
   resist the immunological response of hosts ([34]3). For example,
   extracellular parasites remodel tissues to migrate and evade the immune
   system ([35]4). Similarly, intracellular parasites shape cellular
   processes and remodel host cells to adjust their niche during infection
   ([36]5). The manipulation of these processes and pathways happens
   through molecular interactions that parasites use to their advantage.

   The study of molecular host–parasite interactions is essential to
   understand parasitic infection, local adaptation within the host, and
   pathogenesis. These complex interactions can be described as a network
   ([37]6). Pathogens affect their hosts partly by interacting with host
   proteins, which defines a molecular interplay between the parasite
   survival mechanisms and the host's defense and metabolic systems
   ([38]7). Understanding this molecular crosstalk can provide insights
   into specific interactions that could be targeted to avoid the
   pathological outcomes resulting from the parasitism ([39]8).

   Intra-species protein–protein interactions (PPIs) have been studied in
   depth and there exist large datasets containing experimentally or
   computationally predicted interactions ([40]9, [41]10). However, the
   number of available datasets providing host–pathogen PPIs is limited
   and challenged by the intrinsic difficulties of analyzing
   simultaneously both host and pathogen systems in high-throughput
   experiments ([42]11). Thus, host–pathogen PPIs have mainly been
   predicted computationally using distinct strategies such as approaches
   based on sequence ([43]8, [44]12–[45]15), structure ([46]16, [47]17),
   and gene expression ([48]18). Homology-based prediction is one of the
   most common approaches to predict host–pathogen PPIs. These approaches
   have been extensively used to infer intra-species interactions ([49]10,
   [50]14, [51]19–[52]22) as well as host–pathogen PPIs ([53]13, [54]15,
   [55]17, [56]23) based on the assumption that interactions between
   proteins in one species can be transferred to homolog proteins in
   another species (interologs).

   In this work, we have followed a similar prediction strategy to
   identify common and specific mechanisms of parasitic infection and
   survival across 15 human parasites, namely Trypanosoma brucei,
   Trypanosoma cruzi, Trichinella spiralis, Schistosoma mansoni, Giardia
   lamblia, Plasmodium falciparum, Plasmodium vivax, Plasmodium knowlesi,
   Cryptosporidium hominis, Cryptosporidium parvum, Toxoplasma gondii,
   Leishmania braziliensis, Leishmania mexicana, Leishmania donovani, and
   Leishmania infantum. Our computational prediction approach is based on
   orthology transfer. However, we have constrained the method by (1)
   incorporating only high-confidence intra-species interactions, and
   interactions mined from the scientific literature ([57]10), (2) using
   fined-grained orthology assignments instead of simple sequence
   similarity, and (3) including parasite-specific biological context such
   as lifestyle (uni- or multi-cellular) and tissue infection.

   The objective of these constraints was to reduce the number of falsely
   predicted interactions, increase reliability, and thereby provide a
   better understanding of the parasites' molecular mechanisms of
   interaction with the host. The evaluation of the predicted
   host–parasite PPIs requires repositories of high-throughput
   experimental data that are not yet existent. However, we present here
   extended literature references supporting some of the predicted