Abstract

   Viruses can be spread from one person to another; therefore, they may
   cause disorders in many people, sometimes leading to epidemics and even
   pandemics. New, previously unstudied viruses and some specific mutant
   or recombinant variants of known viruses constantly appear. An example
   is a variant of coronaviruses (CoV) causing severe acute respiratory
   syndrome (SARS), named SARS-CoV-2. Some antiviral drugs, such as
   remdesivir as well as antiretroviral drugs including darunavir,
   lopinavir, and ritonavir are suggested to be effective in treating
   disorders caused by SARS-CoV-2. There are data on the utilization of
   antiretroviral drugs against SARS-CoV-2. Since there are many studies
   aimed at the identification of the molecular mechanisms of human
   immunodeficiency virus type 1 (HIV-1) infection and the development of
   novel therapeutic approaches against HIV-1, we used HIV-1 for our case
   study to identify possible molecular pathways shared by SARS-CoV-2 and
   HIV-1. We applied a text and data mining workflow and identified a list
   of 46 targets, which can be essential for the development of infections
   caused by SARS-CoV-2 and HIV-1. We show that SARS-CoV-2 and HIV-1 share
   some molecular pathways involved in inflammation, immune response, cell
   cycle regulation.

   Keywords: data mining, text mining, virus–host interactions, SARS,
   SARS-CoV-2, HIV-1

1. Introduction

   Studying viruses that infect human beings is vital because viral
   infections can spread within the global population. For several
   centuries, measles, poliomyelitis, rotavirus, human immunodeficiency
   virus, and different respiratory viruses, including influenza,
   parainfluenza, adenoviruses, and coronaviruses, have attacked humanity.
   Such viruses as HIV-1 earlier or SARS-CoV-2 nowadays cause the socially
   significant epidemic or even pandemic diseases [[30]1,[31]2].

   SARS-CoV-2 causes severe acute respiratory syndrome and leads to
   diverse complications, being especially dangerous for people with
   different comorbidities. Starting from the end of January, COronaVIrus
   Disease 2019 (COVID-19) has spread all over the world and become a
   source of fear and sorrow for millions of people. The investigation of
   the interaction of SARS-CoV-2 with the human body is essential because
   such information will contribute to the development of effective
   therapeutic approaches.

   According to the study by Gordon [[32]3], the profile of
   SARS-CoV-2–host interactions is similar to those of involving various
   viruses, such as West Nile virus, human papillomaviruses, Zika, Dengue,
   Ebola viruses, and HIV.

   The dangerous spreading of SARS-CoV-2 has given rise to multiple
   attempts of repurposing of drugs initially developed against other
   viruses, such as hepatitis C, HIV-1, influenza, etc., along with
   host-targeted drugs such as hydroxychloroquine and azithromycin
   [[33]4,[34]5,[35]6,[36]7,[37]8,[38]9].

   Overcoming the challenges of novel viral infections requires a fast and
   effective search for novel therapeutic approaches. A perspective
   strategy is focused on finding proteins involved in the infection
   process caused by both studied and well-known agents using published
   articles and other available information. The identified proteins can
   then be used as pharmaceutical targets. In this study, we aimed at
   evaluating such approach by considering the extensively explored HIV-1
   virus [[39]1] and new dangerous virus SARS-CoV-2 [[40]2].

   There are a few experimental and computational approaches examining the
   possibility of usage of known antiviral drugs, including those employed
   for the treatment of viral hepatitis C and antiretroviral drugs, to
   treat SARS-CoV-2 [[41]7,[42]8,[43]9]. There are several studies
   describing the treatment of SARS-CoV-2 using antiretroviral drugs, such
   as remdesivir [[44]10], lopinavir/ritonavir, and darunavir
   [[45]9,[46]11]. The molecular mechanism of action of
   lopinavir/ritonavir has been discussed [[47]12,[48]13,[49]14]. Their
   application for the treatment of COVID-19 as monotherapies [[50]12] or
   in combination with other drugs [[51]13] has been shown.

   While the mechanisms of action of antiretroviral drugs against COVID-19
   are under investigation, there are data suggesting that SARS-CoV-2 may
   affect the immune system [[52]15]. Lymphopenia can be present in
   patients with COVID-19 [[53]16] and can be associated with disease
   progression [[54]16].

   The analysis of the available information provides (1) data on
   repurposing of antiretroviral drugs against SARS-CoV-2, (2) data
   showing similarity of interaction profiles between SARS-CoV-2 and human
   and HIV-1-human, (3) data indicating common pathophysiological
   mechanisms between SARS-CoV-2 and HIV-1 infections. In addition, there
   are a lot of data on the molecular mechanisms of HIV infection,
   regarding multiple pathways of virus–host interactions and the
   development of novel therapeutic approaches. Text and data mining
   approaches can be helpful for fast and accurately extracting
   information about chemical compounds and their biological activities,
   as well as proteins associated with molecular mechanisms of disease
   development [[55]17,[56]18]. In this study, we applied text and data
   mining approaches to identify possible molecular pathways shared by
   HIV-1 and SARS-CoV-2. Our hypothesis was that the search for similar
   molecular mechanisms shared by SARS-CoV-2 and HIV-1 could shed light on
   novel potential therapeutic approaches for the treatment of COVID-19.

   SARS-CoV-2 and HIV-1 belong to different virus classes, according to
   Baltimore classification [[57]19]. Data on the homology between
   SARS-CoV-2 and HIV-1 are controversial [[58]20]. Besides, they use
   different cells to replicate in the human body. HIV-1 uses the
   receptors CCR5 or CXCR4 to infect CD4 T cells [[59]21,[60]22], while
   SARS-CoV-2 uses ACE2 to infect the cells of heart, lungs, kidneys, and
   gastrointestinal tract [[61]23].

   There are therapeutic strategies to combat HIV-1 infection in addition
   to combination antiretroviral therapy, which is currently used to treat
   HIV-infection [[62]24]. Other approaches to combat HIV-1 infection
   include novel strategies such as disrupting the CCR5 gene in human
   hematopoietic stem/progenitor cells (HSCs) [[63]25] using zinc finger
   nuclease-mediated CCR5 knockout, or in CD4^+ T cells or HSCs using
   CRISPR/Cas9 genome editing [[64]26,[65]27]. Viral RNA genome
   degradation and inhibition of viral gene expression is another
   promising strategy to prevent the replication of SARS-CoV-2 [[66]28]
   and HIV-1 [[67]29]. The search for potent low-molecular-weight drugs
   targeting SARS-CoV-2 is ongoing, but some currently applied treatment
   strategies include the use of some known drugs such as remdesivir,
   chloroquine, umifenovir, and some antiretroviral drugs (darunavir,
   lopinavir in combination with ritonavir) [[68]30].

   The therapeutic effects of medicines applied for the treatment of
   rheumatoid arthritis [[69]31] along with those of some other drugs such
   as antiretroviral drugs (darunavir, lopinavir in combination with
   ritonavir) [[70]32] on SARS-CoV-2 have generated interest in
   identifying similar mechanisms of viral infection shared by SARS-CoV-2
   and HIV-1, in particular their effects on the host’s organism. Another
   important fact is that SARS-CoV-2 can cause serious complications in
   people with comorbidities [[71]11,[72]33,[73]34,[74]35], in particular,
   by causing the so-called cytokine release syndrome [[75]33].

   In this study, we will also discuss the role of the identified
   virus–host interactions in the pathogenesis of SARS-CoV-2 and the
   molecular mechanisms shared by HIV-1 and SARS-CoV-2.

2. Results

   We collected a set of proteins, identified in previous papers, relevant
   to both HIV–host and SARS-CoV-2–host interaction. We searched them
   using different protein and gene databases, including UniProt, KEGG,
   DisGeNet, and Integrity [[76]36,[77]37,[78]38,[79]39]. Some examples of
   the most significant proteins that play a role in both HIV-1 and
   SARS-CoV-2 infections are presented in [80]Table 1.

Table 1.

   Examples of proteins that can have a role in both HIV-1 and severe
   acute respiratory syndrome (SARS)-coronavirus (CoV-2) infections.
   Protein Name UniProt ID ^1 Species ^2 Tissue ^3 Process
   AIP4 [81]Q96J02 Homo sapiens Widely expressed Inflammation
   Beclin 1 [82]Q14457 Homo sapiens ^1 Ubiquitous Autophagy of immune
   cells
   Cathepsin B [83]P07858 Homo sapiens Widely expressed Entry of the virus
   Viral replication (HIV-1)
   Cathepsin L [84]Q5K630 Homo sapiens Widely expressed Entry of the virus
   Complement C3 [85]P01024 Homo sapiens Blood plasma and over 200 tissues
   Immune response
   Inflammation
   Complement activation
   IFITM1 [86]P13164 Homo sapiens Bone and over 200 tissues Immune
   response
   [87]Open in a new tab

   ^1,2,3 UniProt ID, species, tissue are the identifiers of proteins in
   UniProt database.

   For a full list, see [88]Supplementary Materials, Table S1.

   The possible role of the identified proteins in the molecular
   mechanisms of these infections was analyzed.

   Besides the interpretation of the role of individual proteins in
   SARS-CoV-2 and HIV-1 infections, we performed KEGG [[89]37] pathway
   enrichment analysis using genes related to the human proteins found in
   UniProt: we obtained 46 proteins (see Materials and Methods). Then, we
   manually inspected the positions of the genes in the maps of enriched
   pathways and filtered out non-relevant pathways. For example, the FoxO
   signaling pathway was identified by the enrichment analysis, but FOXO
   transcriptional factors were not among those 46 proteins; thus, we
   considered the pathway as non-relevant. Finally, we obtained a list of
   27 pathways ([90]Figure 1):

Figure 1.

   [91]Figure 1
   [92]Open in a new tab

   KEGG pathways enriched in the genes associated with the human proteins
   found by our analysis. Each color represents an individual pathway. The
   size of each box reflects the number of proteins involved in that
   particular pathway. The number of proteins involved in each pathway is
   given in brackets.

   The significant number of signaling pathways related to the immune
   system and the Toll-like receptor signaling pathway included the
   largest number of proteins identified. There were several pathways
   related to various viruses. This means that some of the 46 proteins are
   involved in pathways shared by SARS-CoV-2, HIV-1, and other viruses.
   This result was confirmed by a disease enrichment analysis, which was
   performed based on gene–disease associations using the DisGeNet
   database [[93]38]. Various viral infections and related diseases,
   including SARS-CoV-2 and HIV-1 infections, were among the top enriched
   diseases (see [94]Supplementary Materials, Figure S1).

   To check whether other viruses have similar interactomic profiles to
   that of SARS-CoV-2, we also evaluated texts associated with Dengue
   virus and determined an overlap between the proteins relevant to
   Dengue–host interactions and those relevant to SARS–CoV-2–host
   interactions. We found 32 human proteins, which can be involved in both
   SARS-CoV-2–host and Dengue–host interactions. For them, we prepared a
   list of related pathways. We observed a similar profile of molecular
   pathways for proteins shared by Dengue virus and SARS-CoV-2 to that for
   proteins shared by HIV-1 and SARS-CoV-2 ([95]Figure 2).

Figure 2.

   [96]Figure 2
   [97]Open in a new tab

   KEGG pathways enriched in the genes associated with human proteins
   involved in both SARS-CoV-2–host and Dengue–host interactions. Each
   color represents an individual pathway. The size of each box reflects
   the number of proteins involved in that particular pathway. The number
   of proteins involved in each pathway is given in brackets.

   Nevertheless, we found differences in the number of pathways related to
   the immune system, in particular in the number of proteins associated
   with the Toll-like receptors-related pathway; this is consistent with
   other studies showing the involvement of the immune system in
   SARS-CoV-2 disease progression [[98]15,[99]16]. Also, some of the
   pathways, for instance, necrosis, apoptosis, and autophagy pathways,
   common to SARS-CoV-2–host and HIV-1–host interactions were not found in
   the list of pathways related to SARS-CoV-2–host and Dengue–host
   interactions. This information on proteins involved in the interactions
   between various viruses and the human organism can be helpful for the
   future development of therapeutic strategies against these viruses.

   In addition, by examining the 46 proteins related to both
   SARS-CoV-2-host and HIV-1-host interactions in the Integrity database,
   we identified the main processes associated with targets currently
   validated for the treatment of different pathological conditions and
   diseases. We discuss our results below.

3. Discussion

   By a literature review and a pathway enrichment analysis, we identified
   some key proteins and pathways which may be proposed as essential for
   both SARS-CoV-2 and HIV-1 infections. For instance, the protein AIP4
   may play a role in inflammation, and the interaction of SARS-CoV-2 with
   this protein may downregulate the host’s interferon response [[100]40].
   This would lead to the disruption of the host’s immune response. A
   similar mechanism was reported for HIV-1 [[101]41].

   It has been shown that protein beclin 1 induces autophagy in immune
   cells, which negatively regulates the host’s immune defense mechanisms.
   The possible role of beclin-1 in autophagy as one of the molecular
   mechanisms of viral pathogenesis was discussed for both coronaviruses
   and HIV-1 [[102]42,[103]43]. The inhibition of beclin-1 can help to
   reduce viral replication of MERS-CoV (Middle East respiratory syndrome
   coronavirus) [[104]44]. Despite these findings, it is supposed that the
   viruses can either escape autophagy or use the autophagic machinery for
   their own replication [[105]45]. That is why the strategy of inhibition
   of the proteins included in the autophagic machinery is not considered
   as a promising one to combat viral infections including HIV-1 infection
   and COVID-19 [[106]45].

   The proteins cathepsin B and L, which are endosomal proteases, are
   responsible for viral fusion and the release of the viral genome into
   the cytoplasm of human cells [[107]46]. The dependence of SARS-CoV-2
   entry on cathepsin B was shown in the study by Hoffman et al. [[108]47]
   and discussed in also elsewhere [[109]47,[110]48]. This process was
   shown to be pH-dependent. Divani et al. [[111]48] suggested that the
   Spike (S) protein of the SARS-CoV-2 can be “primed” by cathepsins B and
   L, which is essential for their penetration into the cells. In the
   study by Yoshii. et al. [[112]49], the authors claimed that cathepsin B
   is involved in CD4-independent entry of HIV-1, i.e., this type of entry
   can be used by HIV-1 variants that do not require CD4 for infection
   [[113]49]. There is evidence that an interaction of SARS-CoV-2 with
   cathepsin L is essential for SARS-CoV-2 entry [[114]50]. Selective
   inhibitors of cathepsin L can be considered for the therapy of COVID-19
   [[115]51]. The authors suggest the usage of selective cathepsin L
   inhibitors because active cathepsin S can assist viral release and
   MHC-I- and MHC-II-mediated antigen presentation as well as T-cell
   activation, which is important for protective immunity against
   SARS-CoV-2 and other infections. It is interesting that for
   herpesviruses, the role of cathepsin L in releasing viral particles was
   reported [[116]52], giving rise to the hypothesis that therapeutic
   strategies directed to inhibiting pathways associated with cathepsin L
   can be beneficial for non-specific antiviral activity. Cathepsin B is
   also associated with the release of HIV-1 particles in blood
   monocyte-derived macrophages [[117]53]. Additionally, it was shown that
   secretion of cathepsin B by HIV-infected macrophages leads to neuronal
   internalization and induces the progression of neuronal dysfunctions
   [[118]54,[119]55]. Therefore, cathepsin B expressed in different
   tissues is involved in several different molecular pathways of disease
   progression.

   E3 ubiquitin ligase TRIM25 participates in the regulation of the
   antiviral innate immunity [[120]56,[121]57]. In particular, the
   SARS-CoV nucleocapsid protein, which has over 90% identity to that of
   SARS-CoV-2 [[122]58] was found to bind TRIM25, thereby inhibiting type
   1 interferon production [[123]56]. TRIM25 was also shown to be required
   for the activity of a type I interferon-inducible host factor ZAP,
   which specifically inhibits the replication of certain viruses,
   including HIV-1, Sindbis virus, and Ebola virus [[124]57].

   Some proteins, such as interferon gamma, interferon regulator factor 3,
   interleukin 2, matrix metalloproteinases, are responsible for the
   immune response or inflammation [[125]59,[126]60]. Therefore, these
   proteins should be involved in specific or non-specific antiviral or
   inflammatory response. An increase of interleukins in the blood plasma
   occurs in response to the infection, especially it can be observed
   during viral infections. SARS-CoV-2 affects the immune response,
   leading to a significant increase of chemokines and cytokines. To
   combat the cytokine storm in patients with COVID-19, several
   anti-rheumatic medicines, chloroquine, hydroxychloroquine, as well as
   JAK inhibitors, IL-6 inhibitors, IL-1 inhibitors, anti-TNF-α agents,
   and corticosteroids are considered [[127]59]. The interleukin-6
   receptor antagonist Tocilizumab is described as one of the promising
   drugs to combat the cytokine storm [[128]59]. The development of the
   cytokine storm in the patients infected by HIV-1 was shown [[129]61].

   Interestingly, the role of the c3-complement system activation in
   SARS-CoV-1, SARS-CoV-2-related disorders and HIV-related comorbidities
   was discussed [[130]62,[131]63,[132]64,[133]65]. According to our
   results, complement c3 is involved in both inflammation and immune
   response processes, besides its role as a complement activation. An
   attempt to combat COVID-19 using the compstatin-based complement c3
   inhibitor AMY-101 has been shown recently [[134]65], resulting in a
   good clinical response. The involvement of c3 complement in the mannose
   binding lectin-pathway (MBL) was shown during membrane attack complex
   (MAC) formation, leading to cell lysis and death [[135]66] of
   HIV-infected patients.

   The genes encoding the proteins presented in [136]Table 1 are expressed
   widely or ubiquitously. Therefore, they can participate in similar
   pathways during both SARS-CoV-2 and HIV-1 infections, despite the fact
   that SARS-CoV-2 and HIV-1 use different cell types for their
   replication, which indicates different molecular basis for the
   progression of their associated diseases. Most of the proteins
   identified are Homo sapiens proteins. For them, similar or identical
   molecular mechanisms of pathological processes during HIV-1 and
   COVID-19 can be proposed. This observation confirms the reliability of
   our approach for the extraction of data on proteins involved in the
   interaction of the host with either SARS-CoV-2 or HIV-1.

   When we manually checked about 100 named entities (NE), we found 10 NE;
   and the lexical meaning of them was not related to known proteins but
   it was associated with the biological object, such as a virus, an
   organism, etc. Therefore, about 90% of the proteins found using our
   text and data mining algorithm can be considered as potential targets
   involved in the pathogenesis of both SARS-CoV-2 and HIV-1 infections.

   Pathway enrichment analysis can help to identify a role of some
   signaling pathways in both HIV-1 ad SARS-CoV-2 infection. To perform
   the pathway enrichment analysis, first we manually mapped the initial
   entities, extracted by text mining, to UniProt Accession numbers. This
   process allowed us to collect a list of 46 human proteins. The
   Toll-like receptor-signaling pathway has the highest number of proteins
   related to immune system pathways and found to appear in both
   SARS-CoV-2 and HIV-1 infections ([137]Figure 3).

Figure 3.

   [138]Figure 3
   [139]Open in a new tab

   The Toll-like receptor signaling pathway includes the highest number of
   the proteins identified, related to immune system pathways and
   appearing in both SARS-CoV-2 and HIV-1 infections.

   Toll-like receptors 3 (TLR3) and 7 (TLR7) play an important role in the
   innate immune response against HIV-1 and SARS-CoV-2
   [[140]67,[141]68,[142]69,[143]70,[144]71]. The polymorphism rs3775291
   of the TLR3 receptor is associated with protection against HIV-1
   [[145]67,[146]68]. TLR3 agonists increase the anti-HIV immune response
   induced by vaccination [[147]69]. TLR7 agonists increase the anti-HIV-1
   immune response, the activation of HIV-1 expression, and the inhibition
   of HIV-1 replication, which may help to treat latent infections
   [[148]70]. The role of TLR3 and TLR7 in the pathogenesis of COVID-19 is
   discussed in the study by Sallenave [[149]71].

   Based on 46 human proteins, for which records were found in KEGG, we
   identified the biological processes related to diseases and the
   pathological conditions associated with these proteins using the
   Integrity database. We analyzed the sets of currently validated targets
   for the treatment of certain diseases and pathological conditions.
   According to these results, the majority of proteins are related to
   inflammation, autoimmune disorders and cancer ([150]Figure 4).

Figure 4.

   [151]Figure 4
   [152]Open in a new tab

   The distribution of diseases and pathological conditions by the number
   of protein targets associated with them.

   It seems surprising that most of the found targets of existing drugs
   are not related to infectious diseases but, rather, are associated with
   inflammation, cancer, or immune disorders. However, since we analyzed
   virus–host interactions, we could expect that SARS-CoV-2 and HIV-1
   interaction with the host would be associated with processes like
   inflammation and those that affect the host immune system. The
   association between cancer development and inflammation has been
   discussed in several studies [[153]72,[154]73,[155]74,[156]75].
   Moreover, the role of the targets associated with autoimmune diseases
   in the development of cancer has been described [[157]74]. Based on the
   analysis of the targets strongly related to the found conditions
   according to the Clarivate Analytics Integrity database
   [[158]39]—provided in the [159]Supplementary Materials (Table S2)—we
   identified three main groups of proteins. The first group contains
   interleukins (IL), interferon (IFN), interferon regulatory factor
   (IRF), IL6, IL8, IL12a, IL12b, IL18, IRF3. For these proteins, their
   role in inflammation and the immune response during infection is well
   studied [[160]76]. The second group contains a few proteins, such as
   the Toll-like receptors (TLR) TLR3 and TLR7, whose role in the immune
   response is described above. The third group contains some
   transcription factors and kinases which play a role in various
   molecular processes that are related to the inflammation response. They
   also can be important for the development of cancer. Regarding
   pathological conditions, in general, the results found using Integrity
   are very similar to those found using UniProt and KEGG. Still, an
   interpretation of the results collected from Integrity can help to
   better understand the effectiveness of approved drugs for the treatment
   of SARS-Co-V2 when specific antiviral medicines are not known.

   The obtained results were confirmed by disease enrichment analysis
   based on gene–disease associations using the DisGeNet database. We
   manually analyzed the top 30 enriched disease terms and classified them
   in four groups (see [161]Supplementary Materials, Figure S1). The first
   group includes 12 terms related to viruses and bacteria, e.g., SARS,
   HIV infections, hepatitis B and C, cytomegalovirus infections,
   influenza, tuberculosis, and others. The second group is related to
   inflammatory respiratory diseases: asthma, chronic obstructive airway
   disease, and pneumonia. The third group is related to cancer diseases,
   whereas the fourth group includes all other diseases, which are
   associated with inflammation, e.g., atherosclerosis, systemic lupus
   erythematosus, and rheumatoid arthritis. All 46 proteins appeared to be
   associated with at least one of the selected diseases, and the
   cytokines CXCL8, IL6, TNF were associated with all of them.

   Additionally, we provide the full list of proteins related to the
   development of SARS-CoV-1 infection only in [162]Table S3 of the
   Supplementary Materials.

   Summarizing our findings based on the analysis of the literature and on
   data retrieved from biological databases, we can define groups of
   proteins, which can be essential for both SARS-CoV-2-host and
   HIV-1-host interactions. The first group represents proteins
   responsible for viral entry in human cells, which are the same for
   HIV-1 and SARS-CoV-2 as well as for some other viruses. The second
   group of proteins is associated with the host’s immune response to
   virus invasion of the cells; HIV-1 or SARS-CoV-2 interaction with these
   host proteins will lead to immune suppression. In addition, there are a
   few proteins that were found in just one or two publications but are
   important for processes such as blood coagulation. Third, some viral
   proteins can interact with human proteins in a close manner or share
   some similar structural motifs and can have the same effect on
   protein–host interaction.

4. Materials and Methods

   We applied text and data mining algorithms to extract from publications
   a list of proteins relevant to virus–host interactions common for HIV-1
   and SARS-CoV-2. For this purpose, we performed several stages of text
   processing and data analysis ([163]Figure 5).

Figure 5.

   [164]Figure 5
   [165]Open in a new tab

   General scheme of the data extraction and analysis.

   In the first stage, we collected abstracts and full texts from NCBI
   PubMed and NCBI PubMed Central (PMC) repositories. We used the
   application-programming interface (API) available for PubMed and PMC,
   based on query allowing automated filtering of relevant papers. PubMed
   contains over 27 million articles, and PMC includes several million
   articles; for some of them (the so-called Open Access (OA) subset), it
   is possible to retrieve full texts in a machine-readable format
   (Extensible Markup Language, XML). We used PubMed and the Open Access
   Subset of PMC for our computational experiments. Relevant documents
   were selected based on the following combination of keywords: “HIV-1
   AND virus–host interactions”, “SARS AND virus–host interactions,”
   “SARS-CoV-2 AND virus–host interactions”. The queries were built using
   Python 3.7 scripts. We obtained the list of the identifiers of articles
   in PubMed electronic library (PMIDs) or the identifiers of NCBI PubMed
   Central (PMCIDs). The queries allowed identifying over 200 PMIDs and
   over 6000 PMIDs in the PubMed database relevant to SARS–host and
   SARS-CoV-2-host interactions and HIV1–host interactions, respectively.
   We named these corpora of texts as “SARS-PubMed” and “HIV-PubMed”. Over
   180 PMCIDs were retrieved for SARS–host and SARS-CoV-2–host
   interactions, and over 57,000 PMCIDs were collected for HIV-1–host
   interactions using an automated query (the “SARS-PMC” and “HIV-PMC”
   corpora). We excluded reviews from further consideration. Using PubMed
   API, we collected the texts of over 6200 abstracts of PubMed and over
   38,000 full texts of original articles in XML.

   We used the automatically collected abstracts for the corpora
   “SARS-PubMed” and “HIV-PubMed” and the full texts collected for the
   corpora “SARS-PMC” and “HIV-PMC”. The texts represented in XML format
   were processed to save only the Materials and Methods, Results or
   Results and Discussion sections; the Introduction section was excluded.
   We used the abstracts of the “SARS-HIV-PubMed” corpus in the current
   study because the license agreement restricts the availability of full
   texts. It does not lead to a collection of unbiased corpora (i.e.,
   corpora which are not limited by using several journals). Full texts of
   SARS-PMC and HIV-PMC were used for data extraction.

   The collected corpora of abstracts were processed using Lingpipe-4.1.2
   named entity algorithm trained on GENIA corpus [[166]77]. For
   evaluating the average accuracy of recognition, we applied five-fold
   cross-validation based on the corpus used for training the algorithm
   and an external test set “HIV-RT-inhibitor corpus” containing 250
   abstracts. The average precision for Lingpipe-4.1.2 was 0.84%, the
   recall was 0.79.

   After the extraction of the sets of proteins using all the corpora, we
   automatically compared them and identified overlapping proteins.
   According to our hypothesis, the proteins found in both sets of texts
   that were relevant (a) to HIV-1–host interactions and (b) to SARS-host
   and SARS-CoV-2–host interactions can influence the pathways and
   molecular mechanisms responsible for infection, invasion, and disease
   progression similarly or identically for HIV-1 and SARS-CoV-2.

   The final set of proteins was checked using the UniProt API
   [[167]78,[168]79]. Only proteins found in UniProt and associated with
   Homo Sapiens, SARS-CoV-2, HIV-1 were saved in the list of proteins,
   which we analyzed further. The manual inspection of the proteins
   selected was carried out using NCBI PubMed electronic library and the
   UniProt, Protein Atlas, and Integrity databases.

   To identify signaling pathways, first we manually mapped the initial
   entities, which were extracted by text-mining, to UniProt Accession
   numbers and obtained a list of 46 human proteins. Pathways enriched
   with 46 genes, were identified from the KEGG database [[169]18] using
   the “Enrichr” R package. We selected pathways which included at least 3
   genes from the 46 ones and adjusted the p-value to less than 0.05.
   Visualization of the list of pathways was performed using the “Treemap”
   R package. Additionally, we identified diseases associated with 46
   genes using Enrichr and gene–disease associations using the DisGeNet
   database ([170]https://www.disgenet.org) [[171]38].

5. Conclusions

   We demonstrated the usefulness of text and data mining techniques for
   finding human proteins involved in similar pathological processes
   involving well-known and novel viruses. We used the text and data
   mining techniques to identify possible molecular pathways common to two
   very distinct viruses, HIV-1 and SARS-CoV-2, as a case study. We
   extracted data for proteins related to the molecular mechanisms that
   may be essential for the development of the infections caused by
   SARS-CoV-2 and HIV-1 and demonstrated the usefulness of such
   application. We found a few key proteins that are known to be
   associated with mechanisms related to both HIV-1 and SARS-CoV-2
   infections and to several different processes of pathogenesis, from
   virus entry to evasion of immune response and regulation of
   inflammation in the host. We considered the particular properties and
   molecular pathways that can be involved in different steps of HIV-1 and
   SARS-CoV-2 infections. We provided a few examples of therapies
   targeting some of the identified proteins. We showed that even these
   unrelated viruses can share the same human proteins and regulatory
   pathways, which should be taken into account for the development of
   treatment strategies based on existing drugs. We believe that the
   suggested approach can be used to select known drugs to be applied
   against novel viruses on the basis of shared regulatory pathways and
   protein targets.

Acknowledgments