Protein signature

Abstract

The invention relates to protein signatures, and particularly, although not exclusively, to protein signatures detected in methods and kits for distinguishing between and diagnosing a bacterial and / or viral infection. The invention also relates to the use of a protein signature as a diagnostic or prognostic biomarker for a bacterial and / or viral infection.

Description

[0001] Protein Signature

[0002] The present invention relates to protein signatures, and particularly, although not exclusively, to protein signatures detected in methods and kits for distinguishing between and diagnosing a bacterial and / or viral infection. The invention also relates to the use of a protein signature as a diagnostic or prognostic biomarker for a bacterial and / or viral infection.

[0003] The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7 / 2007-2011) under grant agreement n° 279185 (EUCLIDS). The project leading to this application has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 668303.

[0004] Most febrile children attending healthcare settings have self-resolving viral infections, however, a small minority suffer from bacterial infections which can be life-threatening if untreated. Clinical features do not reliably distinguish between bacterial and viral infections, with confirmed bacterial infections currently identified through culture tests from normally sterile sites. The results can take several days to become available and negative results can be unreliable, hence antibiotics are given on an empirical basis contributing to the spread of antimicrobial resistance (AMR). Conversely, severe bacterial infections can be missed which can have life-threatening consequences. In addition to culture-based diagnostic tests, blood biomarkers, such as C-reactive protein (CRP) and procalcitonin (PCT), are often used as markers of a bacterial infection. Despite their frequent use, CRP and PCT are imperfect biomarkers for distinguishing bacterial from viral infections, as elevated levels of both biomarkers have not only been observed in the plasma of patients with confirmed bacterial infections but also viral infections, including severe acute respiratory syndrome coronavirus 2 (SARS-C0V-2), and many non-infectious conditions. A rapid, accurate, point-of-care test is therefore urgently required for distinguishing between bacterial and viral infections in febrile children. Interrogation of host proteomic profiles obtained from individuals with infectious and inflammatory diseases offers unique insights into disease pathogenesis and can reveal novel protein biomarkers with diagnostic potential. Multiple host protein biomarker candidates for diagnosing febrile illness in children have been identified, with one protein signature (i.e., combination of proteins) proposed and developed into a commercialised point-of- care diagnostic test: MeMed BV™. MeMed BV™ uses host levels of CRP, interferon y- induced protein to (IPio) and TNF-related apoptosis-inducing ligand (TRAIL). The 3- protein signature included in the MeMed BV™ was identified through hypothesis- driven literature searches and targeted screening of biomarker candidates, and developed and optimised for use in populations of all ages. Although it displays promising performance in paediatric populations, there may be alternative protein biomarkers that are superior for diagnosing bacterial and viral infections in children. There is, therefore, still an urgent unmet need for a protein signature that can accurately and reliably distinguish between bacterial and viral infections in febrile children.

[0005] The inventors performed a robust protein biomarker identification process (high- throughput screening phase), using three independent multi-platform proteomic discovery datasets (MS-A, MS-B and SomaScan), using prospectively collected serum and plasma samples from patients with definite bacterial and definite viral infections.

[0006] This identified a total of 431 proteins that were significantly differentially abundant (SDA) between bacterial and viral infections in the SomaScan® dataset (original number of proteins = 1,300). In the MS-A dataset (original number of proteins = 368), 54 proteins were SDA between bacterial and viral infections, and in the MS-B dataset (original number of proteins = 410), 97 proteins were SDA between bacterial and viral infections. The inventors then applied iterative FS-PLS to these large datasets, identifying a total of 35 protein biomarker candidates for quantification in the signature refinement phase, together with 16 additional potential protein biomarkers identified from the literature. The inventors then carried out a signature refinement phase, identifying eight proteins that could distinguish between bacterial and viral infections with surprisingly increased accuracy and reliability.

[0007] Accordingly, in a first aspect of the invention, there is provided a method for distinguishing between a bacterial and viral infection in a subject, comprising detecting, in a sample obtained from the subject, the modulation in protein expression levels of a protein signature comprising at least two proteins selected from a group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, AP0C1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI. Advantageously, as demonstrated in the Examples, the inventors have surprisingly identified a protein signature that can accurately classify 90% of definite bacterial and 82% of definite viral patients. Additionally, the inventors have demonstrated that the protein signature according to the invention significantly outperformed the MeMed BV™ three-protein signature (Figure 5), and, therefore, provides a more accurate and sensitive protein signature for distinguishing between, and diagnosing, bacterial and / or viral infections.

[0008] The invention also provides a method for diagnosing a bacterial and / or viral infection in a subject.

[0009] Preferably, therefore, the method according to the first aspect further comprises a step of diagnosing the subject with a bacterial and / or viral infection. Thus, in a second aspect of the invention, there is provided a method for diagnosing a subject having a bacterial and / or viral infection, the method comprising detecting, in a sample obtained from the subject, the modulation in protein expression levels of a protein signature comprising at least two proteins selected from a group consisting of: SELE, IL18, NCAMl, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR- I, ULBP3, and ZPI.

[0010] The invention also provides a kit for distinguishing between and diagnosing a bacterial and / or viral infection, in a subject.

[0011] Hence, according to a third aspect of the invention, there is provided a kit for distinguishing between a bacterial and viral infection in a subject, the kit comprising means for detecting, in sample obtained from the subject, the modulation in protein expression levels of at least two proteins selected from a group consisting of: SELE, IL18, NCAMl, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI. Preferably, the kit according to the third aspect can also be used for diagnosing a bacterial and / or viral infection in the subject, or for providing a prognosis of the subject’s condition. As described in the Examples, the inventors have demonstrated that the protein signature according to the invention can be used to distinguish between and diagnose a bacterial and / or viral infection.

[0012] Accordingly, in a fourth aspect of the invention, there is provided the use of at least two proteins selected from a protein signature consisting of: SELE, IL18, NCAM1, NGAL,

[0013] IFN-y, LG3BP, SAAl, ANGPT2, A2GL, AACT, AFM, AT3, AP0C1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI, as a diagnostic or prognostic biomarker for a bacterial and / or viral infection.

[0014] The methods according to the invention can aid in the appropriate treatment of patients, for example where it is unclear if the patient is suffering from a bacterial and / or viral infection. This has the advantage of ensuring rapid and appropriate treatment for the bacterial and / or viral infection, and ensures that treatments are only prescribed when the subject is genuinely suffering from such a disease, as opposed to other conditions.

[0015] The rapid identification of a bacterial and / or viral infection would be useful for optimal treatment, and clinical teams require a high degree of confidence in the diagnosis to ensure potentially life-threatening conditions are not missed and to direct empiric treatment and appropriate subsequent investigations.

[0016] Thus, in one embodiment, the method according to the first and / or second aspect comprises a step of administering a therapeutic agent to the subject based on the results of the analysis of the protein signature.

[0017] Accordingly, in a fifth aspect of the invention, there is provided a method of treating a subject suffering from a bacterial and / or viral infection, the method comprising:

[0018] (i) detecting, in a sample obtained from the subject, the modulation in protein expression levels of a protein signature comprising at least two proteins selected from a group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAAl, ANGPT2, A2GL, AACT, AFM, AT3, AP0C1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI, wherein the modulation in protein expression levels of at least two proteins selected from the group consisting of: SELE, IL18, NCAMl, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOC1,

[0019] APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI suggests that the subject suffers from a bacterial and / or viral infection; and (ii) administering, or having administered, to the test subject, a therapeutic agent or putting the test subject on a specialised diet, wherein the therapeutic agent or the specialised diet prevents, reduces or delays progression of the bacterial and / or viral infection. When the subject is diagnosed with a bacterial infection, preferably the method comprises administering an anti-bacterial agent, such as an antibiotic, to the subject.

[0020] When the subject is diagnosed with a viral infection, preferably the method comprises administering an anti-viral agent, an immunomodulatory agent, and / or supportive care where the viral infection is suspected to be self-limiting, to the subject.

[0021] In a sixth aspect of the invention, there is provided a method of detecting, in a subject, the modulation in protein expression levels of a protein signature, the method comprising: (i) obtaining a sample from a subject; and

[0022] (ii) detecting, in the sample, the modulation in protein expression levels of a protein signature comprising at least two proteins selected from a group consisting of: SELE, IL18, NCAMl, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2,

[0023] MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

[0024] The method of the sixth aspect may comprise step (iii) administering, or having administered, to the subject, a therapeutic agent or putting the test subject on a specialised diet, wherein the therapeutic agent or the specialised diet prevents, reduces or delays progression of a bacterial and / or viral infection in the subject. In one preferred embodiment, the protein signature comprises at least three, four, five or six proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN- y, LG3BP, SAAl, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI .

[0025] In another preferred embodiment, the protein signature comprises at least seven, eight, nine, or ten proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAAl, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

[0026] In another preferred embodiment, the protein signature comprises at least 11, 12, 13, 14, 15, 16, 17, 18, 19 or 20 proteins selected from the group consisting of: SELE, IL18, NCAMl, NGAL, IFN-Y, LG3BP, SAAl, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI. In another preferred embodiment, the protein signature comprises at least 21, 22, 23, 24, 25, 26, 27, 28, 29 or 30 proteins selected from the group consisting of: SELE, IL18, NCAMl, NGAL, IFN-Y, LG3BP, SAAl, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

[0027] In another preferred embodiment, the protein signature comprises at least 31, 32, 33, 34, 35, 36, 37 or 38 proteins selected from the group consisting of: SELE, IL18, NCAMl, NGAL, IFN-Y, LG3BP, SAAl, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

[0028] When performing differential abundance analysis, the inventors discovered that LG3BP was the top significantly differentially abundant protein when distinguishing between bacterial infections and viral infections. Accordingly, in a preferred embodiment, the protein signature comprises at least LG3BP.

[0029] In one embodiment, the protein signature comprises at least SELE. In another embodiment, the protein signature comprises at least SELE and IL18. In another embodiment, the protein signature comprises at least SELE, IL18 and NCAM1. In another embodiment, the protein signature comprises at least SELE, IL18, NCAM1 and NGAL. In another embodiment, the protein signature comprises at least SELE, IL18, NCAM1, NGAL and IFN-y. In another embodiment, the protein signature comprises at least SELE, IL18, NCAM1, NGAL, IFN-y and LG3BP.

[0030] In one preferred embodiment, the protein signature comprises at least two proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1 and ANGPT2.

[0031] In another preferred embodiment, the protein signature comprises at least three, four, five, six or seven proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1 and ANGPT2. Preferably, the protein signature comprises all eight proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, SAA1, ANGPT2 and LG3BP.

[0032] In another preferred embodiment, the protein signature consists of: SELE, IL18, NCAM1, NGAL, IFN-y, SAA1, ANGPT2 and LG3BP.

[0033] In another preferred embodiment, the protein signature comprises at least two proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y and LG3BP.

[0034] Preferably, the protein signature comprises at least three, four or five proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y and LG3BP.

[0035] Most preferably, the protein signature comprises all six proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, and LG3BP. In another preferred embodiment, the protein signature consists of: SELE, IL18, NCAM1, NGAL, IFN-y, and LG3BP. Selectin-E (SELE) is an adhesion receptor involved blood leukocyte accumulation at inflammation sites and often isolated from cytokine-stimulated endothelial cells. Neutrophil gelatinase-associated lipocalin (NGAL) can be found in numerous cells (i.e., neutrophil and epithelial cells) and is expressed during inflammation, injury, infection, among other pathologic states. NGAL binds to bacterial siderophores limiting bacterium iron sequestration. Interferon-gamma (IFN-y) is a type II Interferon pleiotropic cytokine critical to both the innate and adaptive immunity. IFN-y functions as an antiviral, antitumor, and immunomodulatory cytokine, and the primary activator of macrophages.

[0036] Interleukin-18 (IL-18) is an IFN-y inducing factor derived from T-cells and natural killer (NK) cells and IL-18 mediates inflammatory pathways via natural killer (NK) cells and macrophages (IL-18 membrane-bound form, sometimes denoted IL-18R).

[0037] Neural Cell Adhesion Molecule 1 (NCAM-i) is a regulator of neurogenesis and the expansion of immune cells (expressed on NK and subsets of T lymphocytes).

[0038] Galectin 3 Binding Protein (LG3BP) plays a role in innate immune responses. LD3BP expression can be induced by viral infections and induced by cytokines such as IFN-y.

[0039] The term “protein signature” as used herein, refers to a set of proteins which when tested together are able to detect and / or discriminate the relevant clinical status. Hence, a protein signature represents a minimal set of proteins which have sufficient discriminatory power to identify and discriminate between a bacterial and viral infection from which the subject is suffering from. Modulation in protein expression levels as used herein means increased or decreased expression of a protein or proteins. Increased expression is intended to refer to a protein which is expressed at higher levels in a diseased or infected patient sample with a bacterial and / or viral infection relative to, for example, a control sample free from a bacterial and / or viral infection. Decreased expression is intended to refer to a protein which is expressed at lower levels in a diseased or infected patient sample with a bacterial and / or viral infection relative to, for example, a control sample free from a bacterial and / or viral infection.

[0040] In one preferred embodiment, the expression of one or more of the following proteins is increased in a subject having a viral infection when compared to a subject having a bacterial infection: IL18, NCAM1, LG3BP, AFM, AT3, CLUS, CNTN5, CO7, FA5, FETUA, IGFBP3, IPSP, ISG15, KAIN, MASP1, MASP2, PLG, and SPP24.

[0041] More preferably, the expression of one or more of the following proteins is increased in a subject having a viral infection when compared to a subject having a bacterial infection: IL18, NCAM1, and LG3BP.

[0042] It will be appreciated, therefore, that the expression of one or more of the following proteins is decreased in a subject having a bacterial infection when compared to a subject having a viral infection: IL18, NCAM1, LG3BP, AFM, AT3, CLUS, CNTN5, CO7, FA5, FETUA, IGFBP3, IPSP, ISG15, KAIN, MASP1, MASP2, PLG, and SPP24.

[0043] More preferably, the expression of one or more of the following proteins is decreased in a subject having a bacterial infection when compared to a subject having a bacterial infection: IL18, NCAM1, and LG3BP.

[0044] In one preferred embodiment, the expression of one or more of the following proteins is increased in a subject having a bacterial infection when compared to a subject having a viral infection: SELE, NGAL, SAA1, ANGPT2, IFN-y, A2GL, AACT, APOC1, APOH, CERU, FBLN3, HRG , LBP, MPIF1, NRP1, SAA2, TIMP1, TNF sR-I, ULBP3, and ZPI.

[0045] More preferably, the expression of one or more of the following proteins is increased in a subject having a bacterial infection when compared to a subject having a viral infection: SELE, NGAL, SAA1, ANGPT2, and IFN-y. It will be appreciated, therefore, that the expression of one or more of the following proteins is decreased in a subject having a viral infection when compared to a subject having a bacterial infection: SELE, NGAL, SAA1, ANGPT2, IFN-y, A2GL, AACT, APOCl, APOH, CERU, FBLN3, HRG , LBP, MPIF1, NRP1, SAA2, TIMP1, TNF sR-I, ULBP3, and ZPI. More preferably, the expression of one or more of the following proteins is decreased in a subject having a viral infection when compared to a subject having a bacterial infection: SELE, NGAL, SAA1, ANGPT2, and IFN-y.

[0046] In one preferred embodiment, the protein expression levels of SELE increase by at least 20%, 40%, 60%, 80%, 100%, 110%, 120%, 130% or 140%, in a subject having a bacterial infection when compared to a subject having a viral infection. More preferably, the protein expression levels of SELE increase by at least 120% in a subject having a bacterial infection when compared to a subject having a viral infection. Most preferably, the protein expression levels of SELE increase by at least 123% in a subject having a bacterial infection when compared to a subject having a viral infection.

[0047] In one preferred embodiment, the protein expression levels of NGAL increase by at least 20%, 40%, 60%, 80%, 100%, 110% or 120%, in a subject having a bacterial infection when compared to a subject having a viral infection. More preferably, the protein expression levels of NGAL increase by at least 100%, or at least 105%, in a subject having a bacterial infection when compared to a subject having a viral infection. Most preferably, the protein expression levels of NGAL increase by at least 106% in a subject having a bacterial infection when compared to a subject having a viral infection. In one preferred embodiment, the protein expression levels of ANGPT2 increase by at least 10%, 20%, 30%, 40%, or 50%, in a subject having a bacterial infection when compared to a subject having a viral infection. More preferably, the protein expression levels of ANGPT2 increase by at least 40%, or at least 45% in a subject having a bacterial infection when compared to a subject having a viral infection. Most preferably, the protein expression levels of ANGPT2 increase by at least 48% in a subject having a bacterial infection when compared to a subject having a viral infection.

[0048] In one preferred embodiment, the protein expression levels of IFN-y increase by at least 10%, 20%, 30%, 40%, or 50%, in a subject having a bacterial infection when compared to a subject having a viral infection. More preferably, the protein expression levels of IFN-y increase by at least 35%, or at least 40%, in a subject having a bacterial infection when compared to a subject having a viral infection. Most preferably, the protein expression levels of IFN-y increase by at least 44% in a subject having a bacterial infection when compared to a subject having a viral infection. In one preferred embodiment, the protein expression levels of SAAt increase by at least 10%, 20%, 30%, 40%, 50%, or 60%, in a subject having a bacterial infection when compared to a subject having a viral infection. More preferably, the protein expression levels of SAAi increase by at least 45%, or at least 50%, in a subject having a bacterial infection when compared to a subject having a viral infection. Most preferably, the protein expression levels of SAAi increase by at least 50% in a subject having a bacterial infection when compared to a subject having a viral infection.

[0049] In one preferred embodiment, the protein expression levels of LG3BP increase by at least 10%, 20%, 30%, 40%, 50%, 60%, or 70%, in a subject having a viral infection when compared to a subject having a bacterial infection. More preferably, the protein expression levels of LG3BP increase by at least 55%, or at least 60%, in a subject having a viral infection when compared to a subject having a bacterial infection. Most preferably, the protein expression levels of LG3BP increase by at least 63% in a subject having a viral infection when compared to a subject having a bacterial infection.

[0050] In one preferred embodiment, the protein expression levels of IL18 increase by at least 2%, 4%, 6%, 8%, 10%, 12%, 14%, or 16%, in a subject having a viral infection when compared to a subject having a bacterial infection. More preferably, the protein expression levels of IL18 increase by at least 8%, or at least 10%, in a subject having a viral infection when compared to a subject having a bacterial infection. Most preferably, the protein expression levels of IL18 increase by at least 11% in a subject having a viral infection when compared to a subject having a bacterial infection. In one preferred embodiment, the protein expression levels of NCAM1 increase by at least 2%, 4%, 6%, 8%, 10%, 12%, 14%, 16%, 18%, 20% or 22% in a subject having a viral infection when compared to a subject having a bacterial infection. More preferably, the protein expression levels of NCAMi increase by at least 18%, or at least 20%, in a subject having a viral infection when compared to a subject having a bacterial infection. Most preferably, the protein expression levels of NCAMi increase by at least 21% in a subject having a viral infection when compared to a subject having a bacterial infection.

[0051] In one embodiment, the methods, kit and / or use of the protein signature according to the invention comprise detecting and / or diagnosing a subject having a bacterial infection only. In another embodiment, the methods, kit and / or use of the protein signature according to the invention comprise detecting and / or diagnosing a subject having a viral infection only. Accordingly, it will be appreciated that the methods, kit and / or use of the protein signature according to the invention comprise diagnosing a subject having a bacterial or viral infection.

[0052] In another embodiment, the methods and / or kit according to the invention comprise detecting and / or diagnosing a subject having a viral infection and a bacterial infection.

[0053] In one preferred embodiment, the bacterial infection is selected from a group consisting of: N. meningitidis, Staphylococcus (e.g. S. aureus), Streptococcus (e.g. S. pneumoniae and S. pyogenes), E. coli, Salmonella spp., Mycoplasma, Campylobacter, M. tuberculosis, C. difficile, Campylobacter, B. pertussis, M. pneumoniae, and Borrelia spp.

[0054] In one preferred embodiment, the viral infection is selected from a group consisting of: Enterovirus, respiratory syncytial virus (RSVj, Rhinovirus, Influenza, Adenovirus, Epstein-Barr virus (EBV), and Measles.

[0055] The subject may be a vertebrate, mammal, or domestic animal. Most preferably, however, the subject is a human being. The subject may be a male or female. The subject may be a child or adult. Preferably, however, the subject is a child. The subject may be under the age of 18, under the age of 15, under the age of 10, or under the age of 5. The subject may be under the age of 4, under the age of 3, under the age of 2, or under the age of 1. Preferably, the subject is under the age of 18. Preferably, the subject is febrile. As used herein, febrile describes a subject with a fever, including symptoms, such as a temperature >38°C. Most preferably, the subject is a febrile child.

[0056] Methods to detect protein expression levels will be well known to the skilled person. For example, protein expression levels may be determined by measuring the abundance of the corresponding DNA or RNA (e.g. mRNA, noncoding RNA (ncRNA), miRNA, tRNA, rRNA, snoRNA, siRNA and piRNA). Alternatively, measuring protein expression at the mRNA level may include measuring levels of cDNA corresponding to mRNA.

[0057] Most preferably, however, protein expression levels are determined by measuring the abundance of the protein.

[0058] Protein expression levels may be detected by a number of ways known to one skilled in the art, for example by Western Blot, immuo-precipitation (IP) or co-IP, immunohistochemistry, bicinchoninic acid (BCA) protein assays, or Luminex assays. Preferably, protein expression levels may be detected using a lateral flow test.

[0059] For example, preferably, immunoassays may be employed to measure protein levels. However, it will be appreciated that non-immuno based assays may be employed, for example, labelling a compound having affinity with a ligand of the protein, and then assaying for the label.

[0060] Protein levels may also be determined with Western Blot analysis. Hence, immunoassays and Western blot analyses may be used to determine the total protein level. Protein concentration may also be detected by enzyme-linked immunosorbent assay (ELISA), fluorometric assay, chemiluminescent assay, or radioimmunoassay analyses.

[0061] The test may be an immunoassay-based test. For instance, labelled antibodies may be used in an immunoassay to evaluate binding of a compound to a protein of interest in the sample. The protein may be isolated and the amount of label bound to it detected.

[0062] The kit may further comprise a label which may be detected. The term “label” can mean a moiety that can be attached to the detection means, or fragment thereof. Moieties can be used, for example, for therapeutic or diagnostic procedures. Therapeutic labels include, for example, moieties that can be attached to an antibody and used to monitor the binding of the antibody to the protein of interest.

[0063] Diagnostic labels include, for example, moieties which can be detected by analytical methods. Analytical methods include, for example, qualitative and quantitative procedures. Qualitative analytical methods include, for example, immunohistochemistry and indirect immunofluorescence. Quantitative analytical methods include, for example, immunoaffinity procedures such as radioimmunoassay, ELISA or FACS analysis. Analytical methods also include both in vitro and in vivo imaging procedures. Specific examples of diagnostic labels that can be detected by analytical means include enzymes, radioisotopes, fluorochromes, chemiluminescent markers, and biotin.

[0064] Protein levels may be measured by double-antibody sandwich ELISA, which will be known to the skilled technician. The ELISA may comprise using a suitable antibody for coating a microtiter plate. Furthermore, the ELISA may comprise using a suitable antibody for detection. The protein of interest, which may be purified from plasma, and which then may be quantified by amino acid analysis, may be used to calibrate a plasma standard using standard techniques known to the skilled technician.

[0065] A label can be attached directly to the antibody, or be attached to the secondary binding agent that specifically binds to the protein of interest. Such a secondary binding agent can be, for example, a secondary antibody. A secondary antibody can be either polyclonal or monoclonal, and of human, rodent or chimeric origin.

[0066] The methods may be carried out in vivo, in vitro or ex vivo. Preferably, however, the method is carried out in vitro.

[0067] The kit of the second aspect may comprise sample extraction means for obtaining the sample from the test subject. The sample extraction means may comprise a needle or syringe or the like. The kit may comprise a sample collection container for receiving the extracted sample, which may be liquid, gaseous or semi-solid.

[0068] The sample is preferably a biological bodily sample taken from the test subject.

[0069] Detecting the modulation of protein expression levels is therefore preferably carried out in vitro. The sample may comprise tissue, blood, plasma, serum, spinal fluid, urine, sweat, saliva, sputum, tears, breast aspirate, prostate fluid, seminal fluid, vaginal fluid, stool, cervical scraping, amniotic fluid, intraocular fluid, mucous, moisture in breath, animal tissue, cell lysates, tumour tissue, hair, skin, buccal scrapings, nails, bone marrow, cartilage, prions, bone powder, ear wax, or combinations thereof. The sample may be a biopsy. Preferably, the sample is blood. Alternatively, the sample may be an ex vivo sample. It will also be appreciated that “fresh” bodily samples may be analysed immediately after they have been taken from a subject. Alternatively, the samples may be frozen and stored. The sample may then be de-frosted and analysed at a later date. The method and / or kit according to the invention may comprise the use of a positive control and / or a negative control against which the modulation in protein expression levels may be compared. For example, a negative control sample is a sample taken from a subject who does not have a bacterial and / or viral infection. A positive control sample is a sample taken from a subject who does have a bacterial and / or viral infection.

[0070] Accordingly, in one embodiment, the method further comprises comparing the protein expression levels of the protein signature with the protein expression levels of the protein signature of a positive and / or negative control. The kit preferably comprises a positive control sample and / or a negative control sample. In one embodiment, an increase in expression levels of a protein compared to the negative control suggests that the subject does have the bacterial and / or viral infection. In another embodiment, a decrease in expression levels of a protein compared to the negative control suggests that the subject does have the bacterial and / or viral infection. All of the features described herein (including any accompanying claims, abstract and drawings), and / or all of the steps of any method or process so disclosed, may be combined with any of the above aspects in any combination, except combinations where at least some of such features and / or steps are mutually exclusive. For a better understanding of the invention, and to show how embodiments of the same may be carried into effect, reference will now be made, by way of example, to the accompanying Figures, in which:-

[0071] Figure 1 illustrates a flowchart depicting an overview of the separate phases carried out and described in the PERFORM (Personalised Risk assessment in Febrile illness to

[0072] Optimise Real-life Management) study, to identify and validate protein biomarkers capable of distinguishing between patients with bacterial and viral infections. First, a robust protein biomarker identification process (high-throughput screening phase) was performed, using three independent multi-platform proteomic discovery datasets. A signature refinement phase was then carried out to identify top candidates using simpler quantification platforms, closer to the type of platform used in a point -of-care diagnostic test. The performance of this signature identified in the signature refinement phase was then tested in a further independent cohort of patients in the signature validation phase. LC-MS / MS: liquid chromatography-tandem mass spectrometry. FS- PLS: forward selection-partial least squares.

[0073] Figure 2 shows the clinical phenotyping algorithm used in the PERFORM project. * = indicates disease groups that could have viral coinfections. § = pathogens normally only detected on mucosal surfaces or diagnosed serologically (M. tuberculosis, B. pertussis, M. pneumoniae, Borrelia species, Campylobacter and Salmonella) can be included in the Definite Bacterial category if clinical presentation is consistent with pathogen detected however may be treated separately for analysis purposes (i.e. , as non-sterile DB).

[0074] Figure 3 shows volcano plots for differential abundance analysis for three separate datasets: the SomaScan®, MS-A, and MS-B datasets. Volcano plots show the log2foldchange (logFC) values and the -logiOBenjamini-Hochberg adj usted p-values for proteins in the SomaScan® (A), MS-A (B), and MS-B (C) cohorts for models contrasting DB to DV samples. Red points = proteins with Benjamini-Hochberg adjusted p-values < 0.05 and absolute logFC values > 1; gold points = proteins with Benjamini-Hochberg adjusted p-values < 0.05; green points = proteins with absolute logFC values > 1, respectively. Black points = not significant (NS).

[0075] Figure 4 shows ROC curves for the signature refinement phase (A) and the signature validation phase (B). A) The ROC curves of the protein signatures identified by running FS-PLS on either Luminex (dark grey line) or Luminex and ELISA proteins (green line) measured in the signature refinement phase. B) The ROC curves of the 5-protein signature identified in the signature refinement phase calculated using original logistic regression weights (blue line), retrained logistic regression weights (orange line), the simple DRS for the 5-protein signature (green line), and the simple DRS for the 6- protein signature (5-protein signature + LG3BP; black line). AUCs (95% confidence intervals) are printed on the plot.

[0076] Figure 5 shows the performance of the 3-protein signature (CRP, IP10, TRAIL) described by Oved et al. Performance was evaluated with (A) and without CRP (B) since CRP was used in the classification of the DV patients included in this study. ROC curves are shown for the 6-protein signature according to the invention (SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP) with (C) and without (D) CRP to enable fair comparison. AUC (95% confidence intervals) are printed on the plots.

[0077] Examples The inventors set out to identify an improved protein signature for distinguishing between and diagnosing bacterial or viral infections in febrile children. In order to do so, the inventors collated multiple high -dimensional proteomic datasets from diverse yet well-phenotyped children with suspected infection who were recruited across multiple European clinical sites, using a previously validated phenotyping algorithm and sampling procedures. The inventors aimed to use these datasets to identify protein biomarkers for diagnosing febrile illness in children and in combination with literature- derived proteins, selected the best combinations of protein biomarkers for distinguishing between bacterial and viral infections. Materials and Methods

[0078] Study design, clinical cohort, ethics statements, case definitions

[0079] The PERFORM (Personalised Risk assessment in Febrile illness to Optimise Real-life Management) study aimed to develop a comprehensive plan for the management of febrile children, including through identifying and validating protein biomarkers capable of distinguishing between patients with bacterial and viral infections. The inventors performed a robust protein biomarker identification process (high- throughput screening phase), using three independent multi-platform proteomic discovery datasets (see Figure 1). The top candidates identified were quantified using simpler quantification platforms, closer to the type of platform used in a point -of-care diagnostic test (signature refinement phase). The signature identified finally underwent validation (validation phase). In the signature refinement phase, a small protein signature for distinguishing between bacterial and viral infections was identified from the previously selected proteins. The performance of this signature was tested in a further independent cohort of patients in the validation phase. At each phase of the study, independent cohorts of patients were used, with no overlap between any patients included in datasets generated at any stage of the study.

[0080] Patient recruitment and case definitions All samples were obtained from patients recruited prospectively into ethics committee- approved studies with written parental informed consent. Patients presenting to healthcare settings experiencing fever (temperature >38.o°C) or a history of fever in the previous 72 hours were recruited. Definite bacterial (DB) and definite viral (DV) samples in the High-Throughput Screening Phase were obtained from patients enrolled in the EUCLIDS study (European Union Childhood Life-Threatening Infectious Disease Study; NRES Committee London - Fulham, 11 / LO / 1982, actively recruiting between July 2012-December 2015), in addition to the Swiss Pediatric Sepsis Study (actively recruiting between September 2011-December 2015), the GENDRES study (Genetic, Vitamin D, and Respiratory Infections Research Network; gendres.org - Ethical Committee of Clinical Investigation of Galicia (CEIC ref 2010 / 015), actively recruiting between July 2009-June 2010) and the PERFORM study (Personalised Risk assessment in Febrile illness to Optimise Real-life Management, perform2020.0rg / ; London - Central Research Ethics Committee, 16 / LO / 1684, actively recruiting between 2016-2021). Independent samples usedin the Signature Refinement and Validation Phases were obtained from the PERFORM study. In the Signature Validation Phase, proteins were measured on DB and DV samples, in addition to samples from patients in other phenotypic groups, including patients with non-sterile definite bacterial infections (non-sterile DB), probable bacterial, bacterial syndrome, probable viral, viral syndrome, and healthy control (HC) groups. HC children were enrolled in the EUCLIDS and PERFORM studies. The phenotypic groups described in Figure 2 were used as the reference groups.

[0081] Case definitions are shown in Figure 2. DB patients are patients in whom an appropriate bacterial pathogen is isolated from a normally sterile site, and clinical presentation is consistent with the pathogen isolated. Detection of virus does not influence assignment. Pathogens normally only detected on mucosal surfaces or diagnosed serologically (M. tuberculosis, B. pertussis, M. pneumoniae, Borrelia species, Campylobacter and Salmonella) can be included if clinical presentation is consistent with pathogen detected but these samples are analysed separately for some analyses (as the non-sterile DB group). DV patients are patients with a viral pathogen isolated that is likely to account for all features of the illness. Patient recruitment was performed in a convenience manner. DB and DV classifications were used as the reference standards as these are gold-standard diagnoses. The number of samples to include at each stage of the study was determined through power calculations, with a minimum of 38 x DB samples identified as required for a test with 95% power.

[0082] High-Throughput Screening Phase Protein quantification

[0083] The high-throughput screening phase involved identifying potential protein biomarker candidates for distinguishing between samples obtained from patients with definite bacterial (DB) and definite viral (DV) infections. Three separate datasets were used for the discovery of protein biomarkers. The “SomaScan®” dataset was generated from serum samples using the multiplexed SomaScan® aptamer-based platform (SomaLogic, Inc.; 1.3K Assay). The remaining two datasets were generated from plasma samples using liquid chromatography tandem mass spectrometry (LC-MS / MS), herein referred to as the “MS-A” and “MS-B” datasets.

[0084] The SomaScan® cohort

[0085] Proteomic profiling for 1,300 human proteins was carried out in plasma samples using a multiplexed assay based on protein capture using modified aptamers (slow off-rate modified aptamers, SOMAmers). Individual protein concentrations are transformed into SOMAmer reagent concentrations which are subsequently measured using a DNA microarray to quantify the original protein’s level. All samples were diluted into three concentrations (40%, 1%, 0.05%) in order to detect proteins with low, medium and high abundances. The serial dilutions in combination with the large dynamic range of each SOMAmer reagent result in a total dynamic range of between -too fM-1 pM for the SomaScan® assay. Samples were split across three plates with DB and DV samples present on each plate in relative proportions.

[0086] The MS-A and MS-B cohorts The proteomic profiles for the MS-A and MS-B cohorts were generated using the same protocols. Prior to proteomic quantification using LC-MS / MS, a top 12 abundant protein depletion spin column (Catalog no. 85165, ThermoFisher Scientific) was used to remove 12 high abundance proteins: ai-acid glycoprotein, ai-antitrypsin, 02- macroglobulin, albumin, apolipoprotein A-I, apolipoprotein A-II, fibrinogen, haptoglobin, IgA, IgG, IgM and transferrin. Ten microliters of plasma were added into each depletion column. Immunodepletion cartridges were rotated at 20 rpm for 1 hr to maximise immunoaffinity binding. The flow-through was collected by spinning the samples at 1000 g for 1 min. To concentrate proteins in flow-through, trichloroacetic acid (TCA) protein precipitation was followed by adding 60 pl of 100% TCA into the Ti2-abundant depleted samples. Six microliters of 2% sodium deoxycholate (DOC) were added into TCA samples and kept at -80 °C for 1 hr. Samples were centrifuged at 13000 g for 30 min at 4 °C and the supernatants were discarded. Eight hundred microliters of cold acetone (100%) were added to each sample, incubated at -20 °C for 1 hr and centrifuged at 13000 g for 30 min. The protein pellet samples were resuspended in 60 pl of 4M urea (pH 7.4 mM Tris). For total protein quantification of the depleted samples, bicinchoninic acid (BCA) protein assays were used (ThermoFisher Scientific).

[0087] Following protein quantification, the proteins were digested at 70° (denaturing condition) by thermostable trypsin using the SMART digestion kit (ThermoFisher Scientific). The purified peptides were analysed by nano-UPLC-MS / MS using a Dionex Ultimate 3000 nano-UPLC with EASY-Spray column (75 pm x 500 mm, 2 pm particle size, Thermo Scientific). A 60 min linear gradient of 2-35% acetonitrile in 5% DMSO / 0.1% Formic Acid was used running at a flow rate of 250 nL / min. MS scans were acquired in the Orbitrap (Fusion Lumos, Thermo Scientific) between 400 and 1500 m / z at a resolution of 120 000, an AGO target of 4.0E5, maximal injection time of 50ms.

[0088] MS / MS were acquired with a fixed duty cycle of 3 seconds (“Top Speed”) and data dependent acquisition (DDA). Precursors between charge state 2+ and 7+ and the intensity threshold of 5.0E3 were selected for higher-energy collisional dissociation (HCD) fragmentation, with collision energy of 35% and an AGC target of 5.0E3. MS / MS spectra were acquired in the ion trap using the rapid scan mode.

[0089] Proteomic profiling was performed at the Discovery Proteomics Facility, Oxford.

[0090] Samples were ordered according to disease group. To avoid bias by system error, the performance of LC-MS / MS (sensitivity of detection, retention time shifting, number of proteins identified) were monitored by running a quality control sample (pool of all individual samples) every 10 samples within batch. A blank in between runs was included cross entire batch which cleans the column and minimises sample carry over. The raw dataset files for MS-A and MS-B were processed separately by MaxQuant (1.6.10.43) based on default settings with matching between runs activated and contaminants included. The FDR cut-off for proteins and peptides was 1% with razor protein FDR enabled and the variable modifications were methionine oxidation and protein N-terminal acetylation. Maximum peptide mass was 4600 Da. Second peptide search was enabled. Relative quantification was performed using the MaxLFQ algorithm.

[0091] Statistical analyses All statistical analyses in this study were performed using the statistical software R (R version 3.6.1). Normalisation and analytical processes were carried out on the three high-throughput discovery proteomic datasets independently due to differences in sample type, study cohort composition, and quantification platform. Limma was used for differential abundance analysis to identify proteins significantly differentially abundant (SDA) between DB and DV. Age and sex were included as covariates for all three datasets, with plate as an additional covariate for the SomaScan® dataset. P- values were adjusted using the Benjamini-Hochberg procedure, with adjusted p-values < 0.05 considered significant. Volcano plots were used to visualise the differential abundance analysis results.

[0092] Feature selection was performed to identify small protein signatures, i.e., combinations of proteins with diagnostic potential. An in-house feature selection method, forward selection-partial least squares (FS-PLS; https: / / github.coni / lachlancoin / fspls.git). was applied to each dataset to identify protein signatures for differentiating between DB and DV infections. Full details of the FS-PLS algorithm and the parameters used are discussed below. FS-PLS was applied across too iterations to each dataset, each time with a different test and training split at a ratio of 70:30. This approach was used to enable identification of the most robust proteins. For each iteration, the signature with the highest area under the receiver operating characteristic (ROC) curve (AUC) in the test dataset was taken forward. The frequency with which each signature and each individual protein were selected across the too iterations was calculated. The “robustness” value was calculated as: rstestssess = shortlist of potential protein biomarkers to take forward to the signature refinement phase was assembled. Multiple inclusion criteria were used to introduce redundancy in case some proteins did not successfully translate across platforms.

[0093] Quality control and pre-processing of proteomic datasets in the High-throughput Screening Phase Quality control steps for the SomaScan® cohort used scale factors returned from the SomaScan® platform to correct for variations in aptamer hybridisation efficiency, inter- and intra-assay variability, variability in the starting quantities of proteins, and plate effects. A hybridisation normalisation scale factor was calculated for each 3 samples using the median values of control SOMAmers. Samples with a hybridisation normalisation scale factor outside of the standard range o.4-2.5 were removed from the analysis. Further batch effect corrections were carried out using COCONUT normalisation and the abundances were log2 transformed.

[0094] For the MS-A and MS-B cohorts, normalisation was performed through the relative quantification step using the MaxLFQ algorithm. MaxLFQ quantifies the relative amount of protein in two or more samples without requiring the proteins to be labelled. Protein groups were removed if they were identified as contaminants or if they were missing in over 90% of samples in each disease group, or in over 90% of all samples to avoid inclusion of proteins that were largely missing. Protein intensities were log2 transformed.

[0095] Following pre-processing, principal component analysis (PCA) was performed on each of the normalised datasets to evaluate the success of quality control (QC) and normalisation in removing the unwanted sources of variation. C-reactive protein (CRP) was removed from each dataset since its levels were originally used to classify viral patients.

[0096] Visualisation of differential abundance analysis

[0097] Differential abundance analysis was performed using Limma and the results were visualised using volcano plots. Volcano plots are scatter plots that show the statistical significance (y-axis) vs. the log2 fold-change (x-axis) for each protein, demonstrating the degree of change between two groups of interest.

[0098] Feature selection with FS-PLS Feature selection was performed using forward selection-partial least squares (FS- PLS). FS-PLS starts by fitting N univariate regression models, where N = the number of features. The maximum likelihood estimation (MLE) function is used to estimate the regression coefficient of each model and a t-test is used to assess the model’s goodness of fit. SVi corresponds to the first variable to be selected. SVi has the highest MLE and smallest p-value. Using singular value decomposition (SVD), FS-PLS projects out the variation explained by SVi enabling it to add additional features that are not correlated to SVi, resulting in a small signature composed of non-correlated features. The following parameters were used: max = 5 (maximum variables to select for the model); p-vahie threshold = 0.001 (p-vahie threshold for selecting a variable / terminating); and beam = 10 (the number of signatures identified in parallel in each iteration). Seed values were used for each dataset to ensure reproducibility of test and training set splits.

[0099] Assembly of shortlist of protein biomarkers from high-throughput screening A shortlist of potential protein biomarkers to take forward to the Signature Refinement

[0100] Phase was assembled. Multiple inclusion criteria were used to introduce redundancy in case some proteins did not successfully translate across platforms. Firstly, proteins identified as SDA between DB and DV in all three datasets were added to the shortlist of potential protein biomarkers as these proteins were consistently altered between bacterial and viral infections regardless of cohort, sample type, and quantification platform. In addition, the top 5 most SDA proteins in each dataset were also added. For each dataset, the proteins included in the signature selected the most frequently by FS- PLS across the too iterations, and the 5 proteins with the highest robustness values overall were added to the shortlist of potential protein biomarkers, with the latter added as they displayed highest stability across subsets of samples.

[0101] Identification of protein biomarker candidates from the literature

[0102] Literature screens were performed, exploring studies that reported biomarkers for diagnosing bacterial and viral infections. Literature searches were performed in December 2017, exploring studies that reported biomarkers for diagnosing bacterial and viral infections. Biomarkers identified were split according to cell type sources, including endothelial cells, T cells, macrophages, neutrophils, leukocytes, and hepatocytes. Searches entailed a thorough screen of all papers in PubMed from January 2005 - December 2017 with a string of key words ‘infection, bacterial or viral, biomarker, plasma or serum’ and additional search terms including ‘biomarker’, ‘cytokine’, ‘chemokine’, ‘growth factor’, and ‘multiplex’ or ‘Luminex’, resulting in approximately 3000 hits.

[0103] Signature Re finement Phase Protein biomarker candidates identified in the high-throughput screening phase were considered for the first round of experimental validation in addition to the proteins identified from the literature. Proteins were quantified using either ELISA or Luminex immunoassays. A protein signature capable of distinguishing between bacterial and viral infections was identified using the iterative, cross-validation FS-PLS approach described above under statistical analysis. Iterative FS-PLS was applied to the proteins measured using Luminex alone (Luminex signature), and then the proteins measured by ELISA in addition to the Luminex proteins (Luminex + ELISA signature). All parameters were the same as described above under statistical analysis, except the number of iterations which was reduced to 25 to reflect the lower number of dimensions. The AUC was calculated for each signature in addition to partial AUCs (pAUCs (16), appendix p4) at 90% sensitivity and specificity, maximal sensitivity and specificity using Youden’s index, and a weighted disease risk score (DRS), calculated through multiplying the abundance values of each protein by their weights / coefficients from FS-PLS. To identify proteins that could distinguish between bacterial and viral infections when CRP is low in bacterial infections, differential abundance analysis was performed between DV samples vs. DB samples with CRP < 6omg / L. Age and sex were included as additional covariates in the model.

[0104] CRP and human neutrophil elastase-ai-antitrypsin (HNE-aiAT) complexes were quantified by ELISA. Mannan-binding lectin serine protease 1 (MASP1; Cloud-Clone Corp., product nr: SEB895HU), interferon-stimulating gene 15 (ISG15; FineTest, product nr. EH1673), secreted phosphoprotein 2 (SPP24; Cusabio, product nr. CSB- EL022604HU), anti-thrombin 3 (AT3; R&D biosystems, product nr. DY1267-05) and serum amyloid Al (SAA1; R&D Biosystems, product nr. DY3019-05) concentrations were also measured by ELISA according to the manufacturer’s protocol. The other selected biomarkers were analysed using a Customized Luminex® human cytokine multiplex panel with custom analytes according to the manufacturer’s protocol (R&D Systems, Inc, Bio-Techne, Minneapolis MN, USA). Protein dilutions were 10-fold across proteins and samples were analysed across 96-well plates. The fluorescence responses and concentrations of analytes were obtained using a Bio-Plex 200 system and the accompanying Bio-Plex Manager Software 6.0. The concentration values and detection limits were determined from standard curves generated from each kit’s standards using the Bio-Plex Software Manager 6.0 (BioRad).

[0105] Extrapolated protein measurements above or below the standard range were excluded to avoid artificially skewing the data. Values were scaled and plate effects from the

[0106] Luminex® assays were removed using ComBat. PCA was performed and samples outside of the 99% confidence ellipse using values from principal components 1 and 2 were removed. Values were log2 transformed. The performance of the signature identified was evaluated through calculating receiver operating characteristic (ROC) curve (AUC) and partial AUCs (pAUC). The pAUC is calculated as the area under the ROC curve in a pre-defined interval. pAUCs were calculated for the regions of the ROC curve within 9O%-1OO% sensitivity and 90%- 100% specificity. The pAUC is beneficial since it describes the performance of the classifier in the most relevant portions of the ROC curve (i.e., regions of high sensitivity and specificity). The abundance values of the proteins in the signature were combined into a single score - the weighted disease risk score (DRS), an adaptation of the DRS described in as follows for each sample i:

[0107] DRSi=valuejix / 3j where j represents the proteins included in the signature and fi represents the model coefficients / weights, obtained from FS-PLS. An example calculation for patient (?) for 4 proteins (k, I, m, n) with values 23.4, 99.4, 46.4, and 74.2, respectively, and weights 0.98, -0.47, 1.23, and -0.12, respectively is as follows: (23.4XO.98)+(99.4X-O.47)+(46.4XI.23)+(74.2X-O.12)=24.382. Signature Validation Phase

[0108] Protein quantification

[0109] The proteins identified in the signature refinement phase as the optimal predictive signature for distinguishing between bacterial and viral infections were taken forward to the signature validation phase by performing further Luminex immunoassays, following the same protocols described previously. Proteins were measured on DB and DV samples, in addition to samples from patients in other phenotypic groups (non- sterile DB, probable bacterial, bacterial syndrome, probable viral, viral syndrome, and healthy controls). The performance of the signature identified in the signature refinement phase in differentiating between DB and DV was evaluated by calculating disease risk scores (DRS) from each sample’s protein abundances. Weighted DRS were calculated using the FS-PLS model weights from the signature refinement phase, as well as retrained model weights from generalized linear models using the signature validation phase data. A simple DRS was also used which entails adding the total abundance of the over-expressed proteins and subtracting the total abundance of the under-expressed proteins. The signature performance was determined through calculating AUCs and pAUCs, and the maximal sensitivity and specificity was calculated using Youden’s index. The signature was used to classify patients, including those in the other disease phenotypic groups (non-sterile DB, probable bacterial, bacterial syndrome, probable viral, viral syndrome), using a DRS threshold corresponding the maximal specificity with sensitivity >90%, to prioritise correct classification of DB infections. Patients were classified in a binary manner (i.e. , DB or DV) with no indeterminate classification. Phenotypic information was available throughout this process. The performance of the signature was compared to the performance of CRP, IPio and TRAIL, a signature described by Oved et al.

[0110] Protein quantification and data pre-processing for the Signature Refinement Phase was the same as outlined in the identification of protein biomarker candidates from the literature. Weighted DRS were used to evaluate the performance of the signature. The ‘original weights protein DRS’ was calculated by multiplying the original model weights determined in the Signature Refinement Phase by the protein abundances measured in the Signature Validation Phase using the equation outlined above. The ‘simple DRS’ was calculated as described in as follows: simple DRSi=^valueki-^valueUml=Onk=O where n and m are the proteins that increase and decrease, respectively, in DB vs. DV. The simple DRS entails adding the total abundance of the over-expressed proteins and subtracting the total abundance of the under-expressed proteins. This approach does not rely on model weights meaning that it is highly translatable and easy to implement in a diagnostic test and is less prone to overfitting. An example calculation for patient (i) for 4 proteins (k, I, m, ri) with proteins k and m decreasing, and proteins I and n increasing, and with values 23.4, 99.4, 46.4, and 74.2, respectively is as follow: (99.4+74.2)-(23.4+46.4)=103.8

[0111] The ‘retrained protein DRS’ was calculated by retraining the model weights using the protein abundances measured in the Signature Validation Phase by using a generalized linear regression model contrasting DB to DV and taking the coefficient estimates as the retrained weights. The DRS was then calculated using the equation used in the identification of protein biomarker candidates from the literature.

[0112] The protein signature identified here was contrasted against the 3-protein signature described in Oved K, Cohen A, Boico 0, Navon R, Friedman T, Etshtein L, et al. A novel host-proteome signature for distinguishing between acute bacterial and viral infections.

[0113] PLoS One. 2oi5;io(3):eoi2ooi2., which is composed of CRP, IP10 and TRAIL. Since CRP was used here in the initial classification of DV samples, the AUC for IP10 and TRAIL combined was contrasted against the AUCs for the signature reported here, using the simple DRS. Then, the performance of IP10 and TRAIL with the addition of CRP was contrasted to the performance of the signature reported here, with CRP added, again using the simple DRS. The simple DRS was calculated as described above with CRP expected to increase in DB, and IPio and TRAIL expected to decrease in DB. ROC models were contrasted against each other using the roc.test function from the pROC package. The performance of the 3-protein signature (CRP, TRAIL, IP10) was also evaluated using a multinomial logistic regression model as used by Oved et al. in their original description of their 3-protein signature.

[0114] Results High-Throughput Screening Phase

[0115] For the discovery of protein biomarkers, the inventors analysed three independent proteomic datasets (Table 1). Two were generated using LC-MS / MS and one was generated by the SomaScan® platform. Table 1: Clinical and laboratory features of patients whose samples were included in the discovery of novel protein biomarkers (high-throughput screening phase). DB = definite bacterial; DV = definite viral; Qi = quartile 1; Q3 = quartile 3.

[0116] A total of 431 proteins were significantly differentially abundant (SDA) (Benjamini- Hochberg adj listed p-vahie <0-05) between bacterial and viral infections in the SomaScan® dataset (original number of proteins = 1,300), with 198 and 233 proteins more abundant and less abundant in bacterial infections, respectively (Figure 3A). In the MS-A dataset (original number of proteins = 368), 54 proteins were SDA between bacterial and viral infections with 20 and 34 proteins more abundant and less abundant, respectively (Figure 3B). In the MS-B dataset (original number of proteins = 410), 97 proteins were SDA between bacterial and viral infections with 28 and 69 proteins more abundant and less abundant in bacterial infections, respectively (Figure 3C). 16 proteins were SDA between bacterial and viral infections in all three datasets with concordant log-fold change directions, and these proteins were added to the shortlist of potential protein biomarker candidates (Table 2). When iterative FS-PLS was applied to the SomaScan® dataset, a 3-protein signature was selected the most frequently (5 / too iterations), composed of ISG15 (robustness: 0-92), TIMP Metallopeptidase Inhibitor 1 (TIMP1; robustness: 0.42), and UL16 Binding Protein 3 (ULBP3; robustness: 0-05). When iterative FS-PLS was applied to the MS-A dataset, a four-protein signature was selected the most frequently (5 / 100 iterations), composed of liposaccharide binding protein (LBP; robustness: 0-55), clusterin (CLUS; robustness: 0-38), apolipoprotein H (APOH; robustness: 0-32), and histidine-rich glycoprotein (HRG; robustness: 0-08). When iterative FS-PLS was applied to the MS-B dataset, a five-protein signature was selected the most frequently (7 / too iterations), composed of AT3 (robustness: 0-95), ceruloplasmin (CERU; robustness: 0-49), secreted phosphoprotein 24 (SPP24; robustness: 0-29), apolipoprotein Cl (AP0C1; robustness: 0-17) and alpha 1-antichymotiypsin (AACT; robustness: 0-09).

[0117] A total of 35 protein biomarker candidates were identified in the high-throughput screening phase and considered for quantification in the signature refinement phase using an independent set of samples, including 18 proteins with elevated levels in bacterial infections, and 17 proteins with elevated levels in viral infections (Table 2). Table 2: The shortlist of protein candidates identified from the high-throughput screening phase. Numbers in brackets following reasons for inclusion show the total number of reasons. SDA = significantly differentially abundant.

[0118] MPIF12C 3-C motif chemokiner, „ . . , SomaScan top 5 SDA P55773 Bacterial

[0119] Literature searches were performed in parallel to the hypothesis-free high-throughput screening phase to identify further protein biomarker candidates. A total of 16 potential protein biomarkers were identified from the literature that also had commercial Luminex or ELISA assays available (Table 3), including ten that increased in bacterial infections and six that decreased in bacterial infections.

[0120] Table 3: Proteins identified as potential biomarkers for distinguishing between bacterial and viral infections in children through literature searches.

[0121] Protein Protein full name Main cell type Disease group with source raised levels

[0122] ANGPT2 Angiopoietin 2 Endothelial cells Bacterial CD163 Scavenger receptor cysteine-rich Macrophages Bacterial type 1 protein M130

[0123] CD25 Interleukin-2 receptor alpha Leukocytes Viral chain

[0124] CRP C-reactive protein Hepatocytes Bacterial CX3CL1 Fractalkine Endothelial cells Viral G-CSF Granulocyte colony-stimulating Monocytes Bacterial factor

[0125] HNE-aATi Human neutrophil elastase-ai- Neutrophils Bacterial antitrypsin

[0126] IFN-gamma Interferon gamma Leukocytes Viral

[0127] IL18 Interleukin 18 Macrophages Bacterial

[0128] IP10 C-X-C motif chemokine 10 T cells, macrophages Viral MPO Myeloperoxidase Neutrophils Bacterial PRTN3 Proteinase 3 Neutrophils Bacterial S100A12 S100-A12 Neutrophils Bacterial inflammation S100A8 S100-A8 Neutrophils Bacterial inflammation

[0129] ST2 Interleukin-i receptor-like 1 Endothelial cells, T Viral cells

[0130] TRAIL Tumor necrosis factor ligand Leukocytes, Viral superfamily member 10 endothelial cells, various tissues

[0131] A total of 13 protein targets had commercial Luminex or ELISA assays available which were taken forward to the signature refinement phase, including five that increased in bacterial infections and eight that increased in viral infections. IFN-y Luminex 3-239 x IO03IL18 Luminex 0-233 (ns) IP1O Luminex 0-071 (ns)

[0132] ISG15* ELISA 0-928 (ns) LBP* Luminex 0-071 (ns) LG3BP* Luminex 1-263 x 10 °'1MASP2* ELISA 0-091 (ns) MPO Luminex 0-174 (ns) NCAM1* Luminex 1-284 x 10 °5 NGAL* Luminex 2-115 x 1011PRTN3 Luminex 2-342 x 10 °’ S100A12 Luminex 3-534 x 10 °6S100A8 Luminex 2-442 x 100SAA1* ELISA 4-385 x io08SELE* Luminex 3-774 x io08SPP24* ELISA 1-108 x 1002ST2 Luminex 4-385 x io08TIMP1* Luminex 1-127 X 10-°« TRAIL Luminex 0-928 (ns)

[0133] Signature Refinement Phase

[0134] Protein levels were measured in an independent set of plasma samples composed of DB (n=88) and DV (0=113) samples (Table 6) using either ELISA or Luminex assays. All samples were plasma samples from patients recruited in the PERFORM study. The aim of the signature refinement phase was to narrow down the list of protein biomarkers identified in the high-throughput screening phase and from the literature screening phase, and to identify the optimal combination of proteins that, when measured using a targeted, simpler platform, display the best performance at distinguishing between bacterial and viral infections.

[0135] The levels of the 13 proteins identified in the high-throughput screening phase as potential biomarkers for distinguishing between DB and DV infections with commercially available ELISA or Luminex immunoassays were evaluated in addition to 16 proteins identified from the literature (Tables 3 and 4). Of the 13 proteins from the high-throughput screening phase, 10 were significantly different between bacterial and viral infections and of the 16 proteins derived from the literature, 11 were significantly different between bacterial and viral infections when levels were compared using Mann-Whitney U tests (Table 4).

[0136] Since the aim of the signature refinement phase was to narrow down the list of 29 protein candidates, iterative FS-PLS was applied to identify small protein signatures for distinguishing DB from DV, with and without the proteins measured using ELISA.

[0137] When FS-PLS was applied to the proteins measured using Luminex assays (proteins = 21), the most frequently selected signature was a 5-protein signature composed of NCAM1, IL18, SELE, NGAL, and IFN-y, which was selected in 7 / 25 iterations (the Luminex signature).

[0138] When FS-PLS was applied to the proteins measured by ELISA assays (except CRP since it was used in the classification of DV samples) in addition to the Luminex proteins (proteins = 28), the most frequently selected signature was a 5-protein signature composed of SELE, IL18, NCAM2, SAA1, ANGPT2, which was selected in 7 / 25 iterations (the Luminex + ELISA signature). When the ROC curves for the Luminex (NCAM1, IL18, SELE, NGAL, IFN-y) and the Luminex + ELISA signatures (SELE, IL18, NCAM1, SAA1, ANGPT2) were compared using the roc.test function from the pROC package, there was no significant difference (p-vahie = 0.844). Despite this, the Luminex signature had a higher overall AUC (89.1% vs. 88.7%; Figure 4A; Table 5), higher maximum sensitivity (90.4% vs. 89.7%), specificity (67% vs. 64.4%) and pAUC when specificity was limited to 9O%-1OO% (6.1% vs. 5.3%; Table 5) and 95%-ioo% (2.4% vs. 2.3%; Table 5). Furthermore, the Luminex signature led to fewer misclassifications than the Luminex + ELISA signature. The Luminex signature was taken forward to validation over the Luminex + ELISA due to its higher sensitivity which was prioritised due to the implications of missing a severe bacterial infection.

[0139] Table 5: The two signatures identified by iterative FS-PLS either using Luminex proteins or Luminex + ELISA proteins. pAUC = partial AUC. Robustness indicates the number of times the protein was included in any signature across the 25 iterations of

[0140] FS-PLS / number of total iterations. Proteins with names italicised and underlined increase and decrease in DB, respectively.

[0141] Differential abundance analysis was performed contrasting DV samples to DB samples with CRP levels < 6omg / L, to identify proteins that can distinguish between bacterial and viral infections when CRP levels are low in DB samples. LG3BP was the top SDA protein, with a Benjamini-Hochberg adj listed p-vahie of 0-013 (log2 fold-change: -

[0142] 0-706). LG3BP was taken forward to validation in addition to the Luminex signature.

[0143] Signature Validation Phase

[0144] Levels of the proteins included in the signature identified in the signature refinement phase were tested on an independent set of plasma samples from patients with DB (0=162) and DV (0=144) infections, in addition to samples in the following phenotypic groups: non-sterile DB (0=31), probable bacterial (0=64), bacterial syndrome (0=2), probable viral (0=91), viral syndrome (0=12) and healthy controls (0=61). All patients were recruited into the PERFORM study. Clinical and laboratory features of the patients are in T able 6.

[0145] Table 6: Clinical and laboratory features of patients whose samples were included in the signature refinement and validation phase. DB = definite bacterial; DV = definite viral; Non-sterile DB = non-sterile definite bacterial; HC = healthy controls; Qi = quartile 1; Q3 = quartile 3. For groups with n=2, medians shown are equivalent to the mean.

[0146] The performance of the 5-protein Luminex signature (Table 5: NCAM1, IL18, SELE, NGAL, IFN-y) in classifying DB and DV samples was tested through calculating the AUC using the original weights estimated from logistic regression models in the signature refinement phase, and separately with retrained model weights estimated from the protein abundance values measured in the signature validation phase. The AUC calculated using the original model weights from the signature refinement phase was 79-7% (95% CI: 74-9%-84-6%; Figure 4B). This improved to 89-2% (95% CI: 85’7%-92-7%; Figure 4B) when retrained model weights were used, however this is likely to be an over-estimation of the signature performance as the model weights are likely to be overfitted to the data.

[0147] The measurements of the proteins included in the signature were combined into a single score for each sample - a simple disease risk score (DRS; appendix p4). The direction (i.e. , whether the proteins are expected to increase or decrease) was determined from the weights calculated by FS-PLS in the signature refinement phase. The AUC using the simple DRS was 87-4% (95% CI: 83-6%-9i-2%; Figure 4B).

[0148] LG3BP was also taken forward from the signature refinement phase for further validation and was combined with the 5-protein signature, leading to a slightly improved AUC of 89-3% (95% CI: 8s-7%-92-9%; Fig. 3B) when the simple DRS was used. When model weights were retrained with LG3BP included in the signature, the 6- protein signature achieved an AUC of 93-6% (95% CI: 90-9%-g6-3%; Fig- 3B). The addition of LG3BP led to statistically significantly differences between the ROC models for the 5-protein and 6-protein signatures calculated using retrained model weights (p- value: 1-5x104) but not for the models calculated using the simple DRS. The addition of

[0149] LG3BP to the signature led to improvements in specificity over the 5-protein signature but not sensitivity. The specificity of the 6-protein signature was 89-6% and 85-4% for ROC models with the retrained weights and simple DRS, respectively. The 6-protein signature (5-protein signature + LG3BP) was used for downstream analyses given its improved specificity in classifying DV samples. The 6-protein signature was applied to the other phenotypic groups in the cohort to determine whether they would be classified as bacterial or viral and was contrasted to the 3- protein signature identified by Oved et al. The 6-protein signature was contrasted against the 3-protein signature with and without CRP, and in both instances, the 6- protein signature outperformed the 3-protein signature with statistically significant differences in AUC (models including CRPp-value: 5-92x104; models excluding CRP p- value: 7-79x108) (Figure s). Discussion

[0150] Novel diagnostic methods for rapid and accurate diagnosis of bacterial and viral infections are required to reduce the amount of antibiotics prescribed unnecessarily. Host protein biomarkers such as CRP and PCT are already extensively used in clinical practise, however they can be unreliable, with elevated levels observed in both bacterial and viral infections. The inventors used high-dimensional proteomic datasets to discover novel protein biomarker targets for distinguishing between bacterial and viral infections in febrile children with the aim of identifying a host protein signature with diagnostic potential. The inventors performed a multi-platform, multi-cohort study to identify and subsequently validate protein biomarkers for differentiating between bacterial and viral infections in febrile children. Using cutting-edge machine learning tools, the inventors have identified a protein signature with high AUCs in distinguishing between confirmed bacterial and viral infections. CRP, TRAIL and IPio have previously been reported as a promising protein signature for diagnosing febrile children, with sensitivity and specificities of 93.7% and 94.2%, respectively. CRP was used in the initial classification of the DV patients reported here (DV patients are required to have CRP < 60 mg / L) as per the PERFORM phenotyping algorithm (Figure 2), meaning direct comparison is challenging. Despite these challenges, the protein signature according to the invention outperformed the Oved et al., 3-protein signature with and without CRP, leading to statistically significant improvements in performance. CRP is an imperfect marker, with elevated CRP in various other infectious and inflammatory conditions, including SARS-C0V-2, influenza, and severe adenovirus.

[0151] The protein signature according to the invention can accurately classify 90% of definite bacterial and 82% of definite viral patients, meaning that it is expected to be robust to coinfections of viral and bacterial pathogens.

[0152] Conclusions

[0153] The majority of febrile children attending healthcare settings have self-resolving viral infections, however, a small minority suffer from life-threatening bacterial infections. Clinical features alone do not reliably distinguish between bacterial and viral infections. Culture-based methods are the gold-standard diagnostic approaches, however, they have various shortcomings including slow turnaround times, low sensitivity and high resource intensity.

[0154] Through a rigorous multi-stage study, using multiple patient cohorts and platforms, the inventors have discovered and subsequently validated various protein biomarkers, resulting in a protein signature that can accurately distinguish between definite bacterial and definite viral infections in febrile children. These proteins could be developed into a blood-based rapid point-of-care diagnostic test for distinguishing between bacterial and viral infections in febrile children, for example as a rule-out test for determining who does not need antibiotics.

Claims

Claims1. A method for distinguishing between a bacterial and viral infection in a subject, comprising detecting, in a sample obtained from the subject, the modulation in protein expression levels of a protein signature comprising at least two proteins selected from a group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOCl, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

2. A method for diagnosing a subject having a bacterial and / or viral infection, the method comprising detecting, in a sample obtained from the subject, the modulation in protein expression levels of a protein signature comprising at least two proteins selected from a group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOCl, APOH, CERU, CLUS, CNTN5, CO7, FA5,FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

3. The method according to either claim 1 or claim 2, wherein the protein signature comprises at least three, four, five or six proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOCl, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

4. The method according to any preceding claim, wherein the protein signature comprises at least seven, eight, nine, or ten proteins selected from the group consisting of: SELE, IL18, NCAMl, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOCl, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1,TNF sR-I, ULBP3, and ZPI.

5. The method according to any preceding claim, wherein:(i) the protein signature comprises at least SELE; (ii) the protein signature comprises at least SELE and IL18;(iii) the protein signature comprises at least SELE, IL18 and NCAMl;(iv) the protein signature comprises at least SELE, IL18, NCAM1 and NGAL;(v) the protein signature comprises at least SELE, IL18, NCAM1, NGAL and IFN-y; or(vi) the protein signature comprises at least SELE, IL18, NCAM1, NGAL, IFN-y and LG3BP.

6. The method according to any preceding claim, wherein the protein signature comprises at least two proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1 and ANGPT2.

7. The method according to any preceding claim, wherein the protein signature comprises at least two proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y and LG3BP.

8. The method according to any preceding claim, wherein the protein signature comprises at least three, four or five proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y and LG3BP.

9. The method according to any preceding claim, wherein the protein signature comprises all six proteins selected from the group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, and LG3BP.

10. The method according to any preceding claim, wherein the expression of one or more of the following proteins is increased in a subject having a viral infection when compared to a subject having a bacterial infection: IL18, NCAM1, LG3BP, AFM, AT3, CLUS, CNTN5, CO7, FA5, FETUA, IGFBP3, IPSP, ISG15, KAIN, MASP1, MASP2, PLG, and SPP24.

11. The method according to any preceding claim, wherein the expression of one or more of the following proteins is increased in a subject having a bacterial infection when compared to a subject having a viral infection: SELE, NGAL, SAA1, ANGPT2, IFN-y, A2GL, AACT, APOC1, APOH, CERU, FBLN3, HRG , LBP, MPIF1, NRP1, SAA2, TIMP1, TNF sR-I, ULBP3, and ZPI.

12. The method according to any preceding claim, wherein:(i) the protein expression levels of SELE increase by at least 20%, 40%, 60%, 80%, 100%, 110%, 120%, 130% or 140%, in a subject having a bacterial infection when compared to a subject having a viral infection;(ii) the protein expression levels of NGAL increase by at least 20%, 40%, 60%, 80%, 100%, 110% or 120%, in a subject having a bacterial infection when compared to a subject having a viral infection;(iii) the protein expression levels of ANGPT2 increase by at least 10%, 20%, 30%, 40%, or 50%, in a subject having a bacterial infection when compared to a subject having a viral infection; (iv) the protein expression levels of IFN-y increase by at least 10%, 20%, 30%,40%, or 50%, in a subject having a bacterial infection when compared to a subject having a viral infection; and / or(v) the protein expression levels of SAAt increase by at least 10%, 20%, 30%, 40%, 50%, or 60%, in a subject having a bacterial infection when compared to a subject having a viral infection.

13. The method according to any preceding claim, wherein:(i) the protein expression levels of LG3BP increase by at least 10%, 20%, 30%, 40%, 50%, 60%, or 70%, in a subject having a viral infection when compared to a subject having a bacterial infection;(ii) the protein expression levels of IL18 increase by at least 2%, 4%, 6%, 8%, 10%, 12%, 14%, or 16%, in a subject having a viral infection when compared to a subject having a bacterial infection; and / or(iii) the protein expression levels of NCAM1 increase by at least 2%, 4%, 6%, 8%, 10%, 12%, 14%, 16%, 18%, 20% or 22% in a subject having a viral infection when compared to a subject having a bacterial infection.

14. The method according to any preceding claim, wherein the bacterial infection is selected from a group consisting of: N. meningitidis, Staphylococcus (e.g. S. aureus), Streptococcus (e.g. S. pneumoniae and S. pyogenes), E. coli, Salmonella spp., Mycoplasma, Campylobacter, M. tuberculosis, C. difficile, Campylobacter, B. pertussis, M. pneumoniae, and Borrelia spp, and / or wherein the viral infection is selected from a group consisting of: Enterovirus, respiratory syncytial virus (RSV), Rhinovirus, Influenza, Adenovirus, Epstein-Barr virus (EBV), and Measles.15- The method according to any preceding claim, wherein the subject is a child, preferably wherein the subject is under the age of 18.

16. The method according to any preceding claim, wherein the subject is febrile.

17. A kit for distinguishing between a bacterial and viral infection in a subject, the kit comprising means for detecting, in sample obtained from the subject, the modulation in protein expression levels of at least two proteins selected from a group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, AP0C1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

18. Use of at least two proteins selected from a protein signature consisting of: SELE, IL18, NCAMl, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR- I, ULBP3, and ZPI, as a diagnostic or prognostic biomarker for a bacterial and / or viral infection.

19. A method of treating a subject suffering from a bacterial and / or viral infection, the method comprising:(i) detecting, in a sample obtained from the subject, the modulation in protein expression levels of a protein signature comprising at least two proteins selected from a group consisting of: SELE, IL18, NCAMl, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOC1, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI, wherein the modulation in protein expression levels of at least two proteins selected from the group consisting of: SELE, IL18, NCAMl, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOC1,APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI suggests that the subject suffers from a bacterial and / or viral infection; and(ii) administering, or having administered, to the test subject, a therapeutic agent or putting the test subject on a specialised diet, wherein the therapeutic agent or the specialised diet prevents, reduces or delays progression of the bacterial and / or viral infection.

20. A method of detecting, in a subject, the modulation in protein expression levels of a protein signature, the method comprising:(i) obtaining a sample from a subject; and(ii) detecting, in the sample, the modulation in protein expression levels of a protein signature comprising at least two proteins selected from a group consisting of: SELE, IL18, NCAM1, NGAL, IFN-y, LG3BP, SAA1, ANGPT2, A2GL, AACT, AFM, AT3, APOCl, APOH, CERU, CLUS, CNTN5, CO7, FA5, FBLN3, FETUA, HRG, IGFBP3, IPSP, ISG15, KAIN, LBP, MASP1, MASP2, MPIF1, NRP1, PLG, SAA2, SPP24, TIMP1, TNF sR-I, ULBP3, and ZPI.

Images