Diagnosis of malaria and arboviruses

To achieve the aims set by the World Health Organisation’s (WHO) Global Technical Strategy for Malaria 2016–2030 which aims to reduce malaria incidence and related mortality by 90% and to eradicate malaria in at least 35 countries by 2030, new strategies to address residual malaria transmission and tools to monitor the results of these strategies are urgently needed [8]. One of the cornerstones for disease control is the availability of good quality vaccines; however, malaria and some arboviruses vaccines are still under development. For example, the only approved malaria vaccine RTS,S (Mosquirix) has a relatively low efficacy and is not recommended by WHO for vaccination of babies between 6 to 12 weeks of age [9]. To achieve the WHO’s 2030 goals of reducing malaria-related mortality by 90%, diagnosis of malaria and mosquito surveillance have been pinpointed as fundamental tools [8]. In addition, to reduce the spread and unprecedented future outbreaks of mosquito-borne diseases, active surveillance of vectors and parasites within human populations is crucial.

Current diagnosis of malaria relies primarily on microscopy methods using Giemsa stained blood smears [10]. However, with a limit of detection of >5 parasites/μL of blood, it requires a well-trained microscopist [11]. Rapid diagnostic tests are also common diagnostic tools for malaria. They are very easy to use and do not require qualified personnel, but their sensitivity and specificity is low in detecting low parasitaemia [12]. Molecular based techniques such as polymerase chain reaction (PCR), quantitative PCR (qPCR), nested PCR, and enzyme-linked immunosorbent assay (ELISA) have also been developed for malaria [13] and arboviruses [14,15]. PCR techniques are gold standards for the diagnosis of arboviruses; however, due to time, cost inefficiencies, and technical expertise required, they are unsuited for large-scale diagnoses particularly during disease outbreaks. For example, the cost of DENV1 antibody ELISA kit is approximately $10.4 USD per sample [16], while a malaria IgG and IgM antibody ELISA kit costs are estimated at $5.5 USD per sample [17]. Additionally, basic laboratory skills are required to perform PCR or ELISA techniques efficiently and correctly.

Vector surveillance of malaria and arboviruses

Vector surveillance involves regular monitoring of mosquito populations to assess the effectiveness of vector control interventions. Surveillance assesses vector survival (age), species diversity, infection status, host preference, and insecticide resistance. These parameters are currently determined using molecular techniques including PCR and qPCR or ELISA [18–20]. Vector survival is the most important determinant of vectorial capacity of mosquito vectors. Mosquito age prediction can be useful in identifying potentially infectious vectors, as pathogens must incubate for a certain period of time within mosquitoes before they can be transmitted to hosts. For example, the female Anopheline mosquito can only transmit Plasmodium parasites after 10 to 12 days following ingestion of an infected blood meal due to the long incubation period required for Plasmodium parasite development within the vector [21]. Consequently, a mosquito population that survives longer that this extrinsic incubation period will be more likely to transmit malaria to susceptible hosts.

Parity dissections to determine whether a mosquito has previously laid eggs or not is the current gold standard technique to determine mosquito age [22]. A related technique which determines the number of dilatations in the ovaries indicates how many times a mosquito has laid eggs [23]. Although these techniques require minimal reagents to operate, they are time consuming and tedious allowing only a small proportion of samples to be dissected at a time which can be an accurate representation of the age composition of a mosquito population.

Raman and infrared spectroscopy

Raman spectroscopy (RS) is a technique that provides chemical fingerprints of molecules by determining their vibrational modes through inelastic scattering of photons known as Raman scattering or Raman effect [24]. During Raman scattering, molecules gain energy from an incident light source. Raman effect is therefore the difference between monochromatic incident and exit radiation. One of the important features that makes RS useful in biological applications is its ability to avoid interference by water molecules. This is due to inability of water to induce Raman scattering. RS can be used as a quantification and identification measure for biological samples. For example, RS has been used to identify molecular compositions in biological samples such as the eyes [25], teeth [26], muscles [27], and nerves [28] and quantify molecular compositions in blood [29–31].

Infrared spectroscopy involves the interaction of infrared radiation with biological samples to produce a diagnostic spectrum. It capitalises on the fact that molecules absorb light at specific frequencies characteristic of their chemical composition [32]. This implies that different biological samples with varying chemical profiles have unique absorption and reflectance properties characteristic of their functional groups and can therefore be quantified as peaks on an infrared spectrum. The infrared portion of the electromagnetic spectrum consist of 3 regions: near-, mid-, and far-infrared. The near-infrared region consists of frequencies that range from 14,000 to 4,000 cm⁻¹ (800 to 2,500 nm wavelength) and is generally used to observe excitation of overtone or harmonic molecular vibrations (Fig 1). The mid-infrared region consists of frequencies that range from 4,000 to 400 cm⁻¹ (2,500 to 25,000 nm wavelength) and is used to study key rotational-vibrational structure (Fig 2). Therefore, mid-infrared wavelengths provide more detailed analyses of a sample. Unlike near-infrared, mid-infrared is invasive and is unsuited for in vivo studies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. An example of an averaged NIRS raw spectra collected from the heads and thoraces of ZIKV-infected (red) and uninfected (blue) Ae. aegypti mosquitoes. Adapted from Fernandes and colleagues [90].

https://doi.org/10.1371/journal.pntd.0009218.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. An example of an averaged MIRS spectra from dried blood infected with malaria parasites (red spectrum) and uninfected blood (blue spectrum).

Adapted from Mwanga and colleagues [70].

https://doi.org/10.1371/journal.pntd.0009218.g002

Both Raman and infrared spectroscopy techniques are rapid and inexpensive techniques compared to molecular and microscopy techniques for similar purposes. Although the initial outlay for a NIR spectrometer can be costly (approximately $40,000 USD), benefits such as minimal sample processing, large-scale applications, and minimal labour can outweigh these initial cost in the long run [33].

Chemometrics/machine learning techniques are usually coupled with spectroscopy techniques to produce diagnostic information required for sample characterisation. Following development of training models, these techniques only require basic computer and spectra collection skills. However, to date, there has not been a comprehensive review of applications of these spectroscopy techniques for mosquito-borne diseases.

We systematically reviewed published evidence from 2009 to 2021 involving the use of Raman and infrared spectroscopy techniques for the diagnosis of malaria parasites and arboviruses and for surveillance of mosquito vectors.

Search strategy

Standard systematic review and meta-analysis (PRISMA) guidelines were applied for this review [34]. We searched PUBMED, MEDILINE, and Web of Science databases for peer-reviewed journal articles published from 2009 to 2021 (January). We manually searched reference lists of included articles to capture relevant articles [35]. To identify articles on the application of RS in the field of mosquito-borne diseases, the following key terms were used: “Raman spectroscopy arboviruses,” “Raman spectroscopy malaria,” “Raman spectroscopy Chikungunya,” “Raman spectroscopy Dengue,” “Raman Spectroscopy mosquitoes,” “Raman spectroscopy Ross River,” and “Raman spectroscopy Zika.” The application of infrared techniques was searched in the same databases with the following key terms: “Infrared spectroscopy arbovirus,” “Infrared spectroscopy malaria,” “Infrared spectroscopy Chikungunya,” “Infrared spectroscopy Dengue,” “Infrared spectroscopy mosquitoes,” “Infrared spectroscopy Ross River,” and “Infrared spectroscopy Zika.” No restrictions were applied to language. EndNote software (Thompson Reuters, Philadelphia, Pennsylvania, United States of America) was used to store articles retrieved from databases which were screened for duplicates. Titles and abstracts were screened by two authors (BG and KC) to identify relevant publications that met the inclusion criteria. Full-text review was applied by one author (BG) to determine the eligibility of articles. Eligible articles were grouped into 3 categories based on the spectroscopy technique used, RS, MIRS, and NIRS.

Inclusion and exclusion criteria

Articles were eligible for inclusion if they demonstrated use of RS, MIRS, and NIRS for the diagnosis, detection, and visualisation of malaria parasites, arboviruses, or surveillance of mosquito vectors. Machine learning articles involving the same sample types were also included in this review. Articles were excluded based on the following criteria: (a) abstract or full paper was not accessible; (b) article does not mention RS, MIRS, or NIRS techniques; (c) article does not mention mosquito-borne diseases; (d) RS, MIRS, or NIRS technique was not used in the main experiments; and (e) conference proceedings, commentaries, grey literature, short communications, or review articles (Fig 3). Articles excluded are indicated in S2 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Search and selection process based on PRISMA framework.

A total of 58 peer-reviewed articles from 2009 to 2021 (January) were reviewed.

https://doi.org/10.1371/journal.pntd.0009218.g003

Data extraction

Eligible articles were subjected to data extraction based on the following criteria: (a) type of spectroscopy technique used; (b) type of sample analysed; (c) method of sample preparation; (d) method of sample analysis; (e) method of data analysis; and (f) result of the experiment based on the spectroscopy technique used (i.e., accuracy defined as the percentage of correct predictions for a sample set, sensitivity defined as the proportion of positive samples that are correctly predicted as positive, and specificity defined as the proportion of negative samples that are predicted as negative).

Results of search strategy

Time trend of journal articles in this systematic review.

There has been an upward trend in the number of peer-reviewed articles published in the field of RS, NIRS, and MIRS from 2009 to 2019, with a more than 2-fold increase in 2015. In 2011, a spike in the number of NIRS and RS articles published was observed. A further increase in articles was observed in 2019 which was mainly associated with the application of MIRS in the field of malaria. A decline in studies from 2020 onwards is likely due to the ongoing COVID-19 pandemic (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The number of articles included in this literature review that are related to RS, MIRS, and NIRS for the diagnosis and surveillance of malaria and arboviruses classified by year of publication.

https://doi.org/10.1371/journal.pntd.0009218.g004

In this review, 27 RS, 11 MIRS, and 20 NIRS studies were included. RS, MIRS, and NIRS studies were first split into 2 groups; malaria and arbovirus studies and each of the 2 groups was further split into articles that focused on diagnostics and vector surveillance (Table 1). All reviewed articles under RS were related to its application for the diagnosis of malaria or arboviruses in whole blood/red blood cells (RBCs)/serum/serum. All MIRS articles were related to its use for the diagnosis of malaria parasite in RBCs and surveillance of malaria vectors. Finally, all but one NIRS article reviewed assessed its use as a vector surveillance tool for malaria and arbovirus vectors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of all articles included in this literature review sorted by type of spectroscopy method and scope of study.

https://doi.org/10.1371/journal.pntd.0009218.t001

Application of RS for the diagnosis of malaria parasites

The application of RS to differentiate ring, trophozoite, and schizont stages of the malaria parasite in human O+ RBCs has been demonstrated. Plasmodium falciparum-infected RBCs at the ring stage showed a characteristic Raman peak at 6,254 nm, while trophozoite and schizont stages had distinct peaks at 13,831 nm. The difference in the characteristic peaks were attributed to the modification of the RBC membrane during the development of the parasite [36]. Another strategy to enhance limit of detection of RS resonance utilised silver nanoparticles synthesised within P. falciparum parasites. A limit of detection of 2.5 parasites/μL for the ring stage of P. falciparum was achieved [37]. In a separate study, both malaria and DENV patient’s whole blood samples were differentiated from healthy samples with 83.3% accuracy with positive likelihood ratios of 0.9529 and 0.9584, respectively [31]. When RS was used to predict Plasmodium vivax infection in human plasma, an accuracy of 86% was achieved [38], whereas 1 parasite/μL of either P. falciparum and P. vivax infections in whole blood could be detected with surface-enhanced Raman spectroscopy (SERS) coupled with a nanostructured gold substrate [39]. A study on how pressure affects the Raman spectra was carried out on synthetic hematin anhydride equivalent to malaria pigment hemozoin. The intensity of RS peaks decreased when synthetic hematin anhydride was subjected to increasing pressure up to 27 kbar above atmospheric pressure [40]. As malaria is more prevalent in tropical countries, therefore, higher temperatures in those countries could lead to increased pressure within blood samples during long-term storage. This can influence the RS diagnostic signature obtained. Further tests using real-world malaria-infected blood samples are required to confirm this phenomenon.

RS has also been used to study both blood and tissue stages of Plasmodium berghei-infected mice. In a study reported by Hobro and colleagues [41], P. berghei infection progression was monitored with a confocal Raman microscope via infected mouse at days 1, 2, 3, 4, and 7 post inoculation. Heme-based changes were observed in mice at a parasitaemia of 0.2% in plasma, and erythrocyte membrane changes were observed on day 4 post inoculation at 3% parasitaemia [41]. In mice infected with the P. berghei ANKA strain, significantly higher heme-based Raman vibrations were observed in the tissue of mice with 5% parasitaemia compared with tissues of uninfected mice indicating possible presence of hemozoin [42].

Several other studies applied RS to visualize the hemozoin pigment produced during malaria infection. An atomic force Raman microscope was used to observe the effect of in vitro treatment procedures on P. falciparum-infected RBCs. Results showed that infected RBCs dried in phosphate buffer solution (PBS) causes localisation of hemichrome at the periphery of RBCs, formaldehyde causes diffusion of haemoglobin into the surrounding areas of the RBCs, while a mixture of formaldehyde (3%) and glutaraldehyde (0.1%) maintained the structural integrity of RBCs [29]. Other studies involving RS visualisation of malaria infection in RBCs include studies that show (a) an increase in hemozoin crystal size over time of infection [43], (b) improvement in visualization of β-hematin crystals with the use of magnetised iron oxide core and silver shell nanoparticles [44], (c) similarity in biochemical compounds found in intracellular and extracellular hemozoin [45], (d) an increase in intravascular heme solubility due to nitric oxide interaction with heme [46], (e) a reduced oxygen-affinity for intracellular haemoglobin [30], and (f) identification of five-coordinate high-spin ferric heme complex in erythrocyte digestive vacuole [47].

Structural analysis of P. falciparum-infected RBCs using RS indicated a lower number of domains arranged in transconformation, an increase in membrane protein and lipids, an increase in deoxygenated haemoglobin, and a decrease in α-helical content with an increase in undefined structures [48]. When resonance RS was used for the structural analysis of iron porphyrins and β-hematin, solid states iron porphyrin [Fe(OEP)]₂O exhibited total symmetric mode v₄ when excited with 782 nm and 830 nm lasers. It was also observed that less supramolecular interactions were present. Based on the difference in excitation and supramolecular interactions, the authors suggested that the intensity of symmetric mode v₄ is strongly affected by C–H–X hydrogen bond interactions [49]. A summary of the studies that applied RS for diagnosis of malaria are shown in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. A summary of articles reviewed that applied RS for the diagnosis of malaria parasite.

https://doi.org/10.1371/journal.pntd.0009218.t002

Application of RS for the diagnosis of arboviruses

A sensitivity of 97.38% and a specificity of 86.18% were achieved when RS was used to detect DENV in blood plasma relative to ELISA testing for nonstructural protein 1, immunoglobulin M (IgM), and immunoglobulin G (IgG) [50]. However, a lower predictive accuracy of 66% and 47% was observed when RS was used to detect DENV-infected blood sera relative to IgG and IgM ELISA tests, respectively. The predictive accuracy of DENV based on IgG antibodies was lower than the accuracy based on IgM probably because IgM is generated at the onset of the infection before IgG. Overall, low accuracies were due to high false negative results [51]. Elevated lactate concentration in human blood sera was observed in DENV-infected patients possibly due to impaired function of body organs [52]. When support vector machine (SVM) learning models coupled with polynomial kernel order 1 were applied, DENV was predicted with a predictive accuracy of 85% and a sensitivity of 73% [53]. In a recent study, RS was used to differentiate between bacteria (Salmonella Typhi) and virus (DENV) infections in human blood serum, with 12 distinct Raman bands linked to typhoid-infected samples and 4 to DENV [54].

Multiple studies have been conducted to determine the limit of detection of RS for identifying arbovirus-related antigens, with emphasis on Rift Valley fever virus (RVFV) and West Nile virus (WNV) using SERS coupled with gold and silver nanotags [55–58]. When gold paramagnetic nanoparticles were used, the limit of detection for WNV-specific target DNA sequence was 10 pM [57]. The same methodology was used to identify the limit of detection of RVFV (20 nM) and WNV (100 nM) based on their RNA sequences [58]. RS and gold paramagnetic nanoparticles tags were also used to simultaneously detect WNV and RVFV where detection limits of 5 fg/mL and approximately 25 pg/mL were achieved when viruses were suspended in PBS or PBS with fetal bovine serum, respectively [56]. However, when a polyacrylic acid layer was applied, a reduction in background noise was observed and a limit of detection of 10 pg/mL for WNV, RVFV, and Yersinia pestis antigens was achieved [55].

Detection limits as low as 10 plaque-forming units (PFU)/mL were achieved for both DENV and WNV when SERS coupled with a bioconjugated gold nanoparticle was used [59]. Whereas a limit of detection of 7.67 ng/mL for DENV and 0.72 ng/mL for ZIKV was achieved by coupling RS to a lateral flow assay [60]. These detection limits are lower relative to ELISA detection limits of 120 ng/mL for ZIKV, <1 ng/mL for DENV, and 61 PFU/mL for WNV [61–63]. A summary of the studies that applied RS for the diagnosis of arboviruses are shown in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. A summary of the articles reviewed that applied RS for the diagnosis of arboviruses.

https://doi.org/10.1371/journal.pntd.0009218.t003

Application of MIRS for the diagnosis of malaria parasites

Asexual stages of P. falciparum could be differentiated with Synchrotron Fourier transform infrared microspectroscopy and artificial neural network (ANN) based on the 2,800 to 3,100 cm⁻¹ and to 1,000 to 1,800 cm⁻¹ MIR regions with an accuracy of 100% for all stages tested (rings, trophozoites, and schizonts) [64]. Stages of P. falciparum in human O+ RBCs were differentiated with attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy where rings, trophozoites, and gametocytes were distinct within 1,000 to 3,100 cm⁻¹ regions. The study also identified that the limit of detection was <1 parasite/μL for the ring stage blood samples [65].

When MIRS was used for the detection of P. falciparum in human whole blood with parasitaemia ranging from 0% to 5%, a sensitivity of 70% and a specificity 98% was achieved [66]. Moreover, high-resolution infrared images of single cells infected with P. falciparum, changes in MIRS spectra region relating to increases in amide A band (consisting of N–H stretching modes of protein and C–H stretching region of lipids), and unsaturated fatty acids in infected cells could be identified [67]. Focal plane array-Fourier transform infrared (FPA-FTIR) imaging spectroscopy identified P. falciparum blood stages at a single-cell level [68] and a high-resolution FTIR could detect single malaria parasite-infected erythrocytes [67].

The effect of 3 different anticoagulants including sodium citrate, potassium ethylenediaminetetraacetic acid, and lithium heparin on plasma and whole blood in aqueous and dry phase on ATR-FTIR spectral signatures was tested. It was found that anticoagulants heavily influenced the spectra of dry blood samples compared to wet samples. Of the 3 anticoagulants tested, lithium heparin affected the mid-infrared spectra the least [69]. Findings from 2 most recent studies indicate MIRS can detect P. falciparum field-collected human blood spots on filter paper. Mwanga and colleagues identified P. falciparum with FTIR from field samples collected in Tanzania where an accuracy of 92%, sensitivity of 92.8%, and specificity of 91.7% in comparison to PCR findings was achieved [70]. The second study was done with P. falciparum samples collected in Thailand where a sensitivity of 92% and a specificity of 97% in comparison to PCR was observed [71]. A summary of the studies that applied MIR for diagnosis of malaria are indicated in Table 4, and an example of a MIR spectra for malaria infected and uninfected red blood cells is shown in Fig 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. A summary of articles reviewed that applied MIRS for diagnosis of malaria parasites.

https://doi.org/10.1371/journal.pntd.0009218.t004

Application of MIRS for surveillance of malaria and arbovirus vectors

Highly variable accuracies were observed when MIRS was used to predict the age of laboratory-reared Anopheles arabiensis and Anopheles gambiae that ranged between 1 to 15 days old. Predictive accuracies of 15% to 97% and 10% to 100% were attained for An. gambiae and An. arabiensis mosquito species, respectively. Lower predictive accuracies were observed for middle age mosquitoes within 3 to 11 days old age group compared to 1 or 15 days old mosquitoes for both species [72]. The same technique also differentiated between field-collected An. arabiensis and An. gambiae with an accuracy of 82.6% [72]. A separate study used ATR-FTIR spectroscopy to predict the age of laboratory-reared wMel-infected Ae. aegypti with an accuracy of 95% to 97% and to detect Wolbachia infections in Ae. aegypti field mosquitoes with an accuracy of 90% compared with that of PCR results. However, higher predictive accuracies of 95% to 97% were observed when mosquitoes were 2 and 10 days old. A significant difference in biochemical components between male and female was also identified by ATR-FTIR spectroscopy and the technique differentiated the 2 groups with a specificity and sensitivity of 95% to 100% [73].

Identification of the origin of a mosquito blood meal is crucial for the assessment of their host preference. Furthermore, upscaling MIRS into field studies will require models that are robust enough to accurately predict mosquitoes with various abdominal statuses including the source of their blood and blood digestion stages. MIRS has been used to successfully differentiate laboratory-reared An. arabiensis fed on goat, bovine, chicken, and human blood meals with predictive accuracies of 96%, 97%, 100%, and 100%, respectively [74]. A summary of the studies that applied MIR for surveillance of malaria and arbovirus vectors is shown in Table 5.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. A summary of articles reviewed that applied MIRS for surveillance of malaria and arbovirus vectors.

https://doi.org/10.1371/journal.pntd.0009218.t005

Application of Visible-NIRS for malaria and arbovirus vector surveillance

All studies on surveillance of mosquito vectors were conducted using a Labspec NIR spectrometer (Malvern Panalytical, Malvern, United Kingdom) whose wavenumber range is 4,000 to 28,571 cm⁻¹. Most studies applied partial least squares (PLS) regression for data analysis to predict the age, infection, and species of mosquitoes of malaria and arbovirus-transmitting mosquitoes (Table 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. A summary of articles reviewed that applied Visible-NIRS for diagnosis and surveillance of malaria and arbovirus vectors.

All studies on surveillance of mosquito vectors were conducted using a LabSpec NIR spectrometer (Malvern Panalytical, Malvern, UK).

https://doi.org/10.1371/journal.pntd.0009218.t006

In the past 12 years, NIRS has been used as an alternative strategy for age and species determination for malaria and arbovirus vectors in a range of studies under varying environmental conditions. Initially, NIRS’ potential for predicting the age and species was demonstrated on laboratory-reared mosquitoes where it was reported to be accurate for predicting the age of those mosquitoes into ±3 days of their actual age or into ≤ 7 days ≥ old age group and for differentiating An. gambiae from An. arabiensis with >90% accuracy [75]. Subsequent studies using An. gambiae and An. arabiensis reared in a semi-field system reported similar accuracies for age and species prediction [33,76]. More recent studies have indicated that the chronological age of An. gambiae and An. arabiensis can be improved using alternative machine learning techniques such as ANN as opposed to PLS regression analysis used in previous studies [77,78].

NIRS spectral features of laboratory-reared and wild-type An. arabiensis mosquitoes were shown to be identical regardless of the environment and diet of the mosquito (i.e., field-caught and laboratory-reared mosquitoes). Based on these findings, NIRS prediction models developed from laboratory-reared mosquitoes with known age could be possibly relied upon for predicting the age of wild mosquito populations with unknown age [79]. Two studies have demonstrated the potential of NIRS to predict the age of wild An. arabiensis and An. gambiae mosquitoes. Krajacich and colleagues [80] reported age prediction accuracy of 73.5% to 97% for wild and 69.6% for semi-field mosquitoes. This accuracy has recently been improved using an autoencoder and ANN to predict the parity status of field mosquitoes [78]. Findings from 2 other studies indicated that NIRS could differentiate field-caught An. gambiae and An. arabiensis with a predictive accuracy of 90% [33,75]. NIRS has also been used to predict the age of laboratory-reared Ae. aegypti [81] and Ae. albopictus [82] with or without Wolbachia with similar accuracies as those recorded for An. gambiae and An. arabiensis. However, when NIRS was used to predict Ae. albopictus mosquitoes reared from wild pupae using a model developed from laboratory-reared mosquitoes, young and old mosquitoes could not be differentiated [83]. Based on the authors’ description of their experimental design, the inability to predict the age of mosquitoes collected from wild pupae is most likely due to a weak predictive model that failed to capture the heterogeneity of the wild population including variation in the larval diet. Alternatively, the small sample size used for model development was not robust [83]. A previous study indicated that Anopheles mosquitoes reared from wild pupae could be predicted accurately if models were developed from a similar mosquito population and neither species type nor exposure to pyrethroids affected the ability of NIRS to predict their age [76]. However, larval and adult diets have been previously shown to have an influence on age-related spectral signatures [84].

Mosquitoes are commonly stored in various preservatives prior to spectral collection. Several studies have demonstrated the effect of a range of preservation techniques on the predictive accuracy of NIRS for age and species of An. gambiae and An. arabiensis. When RNAlater was used as a preservative for 1 to 3 weeks, the prediction accuracy was relatively higher (90%) for An. gambiae s.s and An. arabiensis compared to the accuracy of 83% for freshly scanned mosquitoes [85]. However, there appeared to be a decline in accuracy from 86% to 75.6% when mosquitoes were preserved for a longer period of time (50 to 62 days) [86]. Similarly, a light decrease (on average as 3.6%) in accuracy when differentiating An. arabiensis from An. gambiae were observed after 50 weeks of storage in silica gel [87]. For storage up to 4 weeks, RNAlater at 4°C, refrigeration at 4°C and silica gel are the recommended options for age and species prediction [86,87]. The fact that the prediction accuracy of samples stored in silica gel is comparable to the accuracy of fresh samples is encouraging as it means vector control programs would not be required to modify their current collection and storage protocols to adopt the NIRS technique. However, although silica gel is cost-effective, maintaining desiccation for a long period of time is a challenge. Due to its high cost, RNAlater is generally recommended for small-scale studies that require RNA extraction to stabilise RNA in the samples prior to analysis. Samples in RNAlater can stay at room temperature for a maximum of 2 weeks, hence RNAlater could be an alternative to silica gel for field work where access to a fridge is limited.

NIRS has been used to identify mosquitoes infected with various pathogens such as P. berghei, P. falciparum, Wolbachia, CHIKV, and ZIKV. A prediction accuracy of 72% was achieved when P. berghei-infected Anopheles stephensi were differentiated from uninfected mosquitoes [88]. P. falciparum was detected in laboratory-reared An. gambiae, with accuracies of 88% for oocyst stage and 95% for sporozoite stage (14 dpi). This predictive accuracy positively correlated with the concentration of the parasite within the mosquito [89]. ZIKV in laboratory-reared Ae. aegypti mosquitoes was predicted with an overall accuracy of 97.3% in the heads/thoraces and 88.8% in abdomens in comparison to RT-qPCR [90] and wMel-infected Ae. aegypti mosquitoes were predicted with accuracies of 92% for female and 89% for male mosquitoes [91]. Similarly, the presence of wMelPop in Ae. aegypti females and males was predicted with accuracies of 96% and 87.5%, respectively [91]. Furthermore, NIRS could differentiate between wMel and wMelPop-transinfected mosquitoes with predictive accuracies of 96.6% for females and 84.5% for males [91]. Overall, NIRS detected Wolbachia in females more accurately than in male mosquitoes and in wMelPop more accurately than wMel probably based on concentration levels. Lastly, a recently published article provides evidence that NIRS can also detect the presence of Wolbachia, ZIKV, and Chikungunya viruses in mosquitoes 7 days post their death [92]. A summary of the studies that applied NIRS for surveillance of malaria and arbovirus vectors is shown in Table 6, and an example of NIRS raw spectra of ZIKV infected and uninfected female Ae. aegypti is shown in Fig 1.

Application of Visible-NIRS for diagnosis of malaria parasites

Only one study has applied NIRS to detect P. berghei in the whole blood of mice infected with rodent malaria. The study compared NIRS spectra of 6 P. berghei-infected mice and 6 uninfected mice. A characteristic peak at 650 nm related to P. berghei infection increased in intensity with rising parasitaemia (R² value = 0.68) [93]. The study referenced has been added to Table 6.

Knowledge gaps in the application of Raman and infrared spectroscopy for malaria and arboviruses

Although several studies have demonstrated the potential of RS for diagnosis of both malaria and arboviruses in laboratory settings (Table 2), the validation of RS under real-world conditions is an area that has not been fully investigated. Moreover, no studies were identified that have assessed the potential of RS for surveillance and characterisation of mosquito vectors. The use of MIRS for diagnosis of malaria has been recently demonstrated in the laboratory by several studies and in the field by 2 studies [70,71], whereas its application as surveillance tool for malaria vectors has only been demonstrated by 3 studies in the laboratory [72–74]. Future research should assess the potential of MIRS for the diagnosis of arboviruses in humans and validate its feasibility under field conditions for both malaria and arboviruses. NIRS has been used in several studies for the surveillance and characterisation of mosquito vectors into age groups, species identity, and infection status. However, most of the studies were conducted on laboratory samples. Only 2 studies reported that NIRS can differentiate field-collected An. gambiae from An. arabiensis [33,75] and parous from nulliparous malaria vectors [78,80]. Further work is required to demonstrate the full potential of NIRS in the field and to validate it against gold standard techniques. Finally, only 1 study demonstrated that NIRS can detect Plasmodium in mice blood [93] opening an opportunity to investigate its diagnostic capacity for malaria and arboviruses in human tissues.