Rapid classification of epidemiologically relevant age categories of the malaria vector, Anopheles funestus

Rapid classification of epidemiologically relevant age categories of the malaria vector, Anopheles funestus | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Rapid classification of epidemiologically relevant age categories of the malaria vector, Anopheles funestus Emmanuel P. Mwanga, Doreen J. Siria, Issa H. Mshani, Sophia H. Mwinyi, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3834184/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 18 Mar, 2024 Read the published version in Parasites & Vectors → Version 1 posted 9 You are reading this latest preprint version Abstract Background Accurately determining the age and survival probabilities of adult mosquitoes is crucial for understanding parasite transmission, evaluating the effectiveness of control interventions and assessing disease risk in communities. This study was aimed to demonstrating rapid identification of epidemiologically relevant age categories of Anopheles funestus , a major Afro-tropical malaria vector, through the innovative combination of infrared spectroscopy and machine learning, instead of the cumbersome practice of dissecting mosquito ovaries to estimate age based on parity status. Methods An. funestus larvae were collected in rural south-Eastern Tanzania and reared in the insectary. Emerging adult females were sorted by age (1–16 day-olds) and preserved using silica gel. PCR confirmation was conducted using DNA extracted from mosquito legs to verify the presence of An. funestus and eliminate undesired mosquitoes. Mid-infrared spectra were obtained by scanning the heads and thoraces of the mosquitoes using an ATR FT-IR spectrometer. The spectra (N = 2084) were divided into two epidemiologically relevant age groups: 1–9 days (young, non-infectious) and 10–16 days (old, potentially infectious). The dimensionality of the spectra was reduced using principal component analysis, then a set of machine learning and multi-layer perceptron (MLP) models were trained using the spectra to predict the mosquito age categories. Results The best performing model, XGBoost, achieved an overall accuracy of 87%, with classification accuracies of 89% for young and 84% for old An. funestus . When the most important spectral features influencing the model performance were selected to train a new model, the overall accuracy increased slightly to 89%. The MLP model, utilising the significant spectral features, achieved higher classification accuracies of 95% and 94% for the young and old An. funestus , respectively. After dimensionality reduction, the MLP achieved 93% accuracy for both age categories. Conclusion This study shows how machine learning can quickly classify epidemiologically relevant age groups of An. funestus based on their mid-infrared spectra. Having been previously applied to An. gambiae, An. arabiensis and An. coluzzii , this demonstration on An. funestus underscore the potential of this low-cost, reagent-free technique for widespread use on all the major Afro-tropical malaria vectors. Future research should demonstrate how such machine-derived age classifications in field collected mosquitoes correlate with malaria in human populations. Malaria Anopheles funestus deep learning machine learning Ifakara health institute mid-infrared Spectroscopy Figures Figure 1 Figure 2 Figure 3 Background Despite significant investments in malaria control and research, there were still an estimated 249 million malaria cases and 619,000 deaths in 2021 globally, nearly all of which in sub-Saharan Africa [ 1 ]. Other than the poor economic conditions and weak health systems, the continued high burden of malaria in Africa is attributable to key biological threats, notably malaria resistance to drugs [ 2 – 4 ], vector resistance to insecticides [ 5 , 6 ], increasing occurrence of malaria parasites evading detection by rapid diagnostic tests [ 7 – 11 ], and disruptions from major disease outbreaks such as Ebola and COVID-19 [ 12 – 14 ]. Effective vector control, primarily with insecticide treated nets (ITNs) and indoor residual spraying (IRS), has been the most important component of malaria control in Africa [ 15 ]. However, its continued effectiveness requires active innovation to address the current threats, and improved understanding of the major vector species in different settings. Anopheles funestus is one of the four main malaria vector species in sub-Saharan Africa, the others being An. gambiae, An. arabiensis and An. coluzzii , and also one of the most widespread [ 16 – 19 ]. An. funestus is particularly important in East and Southern Africa, where it is becoming the dominant malaria vector. For example, in parts of Tanzania, An. funestus is reported to be responsible for 86–97% of all new malaria infections [ 17 , 20 – 22 ]. Its dominance is due to multiple factors, including i) being highly anthropophilic, and thus preferring to bite humans over other vertebrates [ 17 , 23 ], ii) being highly endophilic, i.e. preferring to bite inside human dwellings than outside [ 24 ], iii) having significantly higher survival rates than other species [ 25 ], iv) being resistant to commonly used insecticides [ 17 , 18 , 26 ] and v) preferentially breeding in perennial habitats with year-round productivity [ 27 ]. Given its importance and dominance in malaria transmission systems, vector surveillance programs in the respective countries should be designed with special attention to this vector species. Besides evaluating biting densities and Plasmodium infection rates, accurately determining the age and survival of An. funestus is crucial for monitoring transmission dynamics and assessing the effectiveness of vector control interventions such as ITNs and IRS. Dissection of mosquito ovaries is still the main entomological technique for estimating the age of vector populations [ 28 ]. The dissections are usually performed under light microscopes to assess the reproductive history, specifically the parity status, of the mosquitoes. This involves observing whether the ovaries contain coiled tracheolar skeins (indicating non-parous mosquitoes) or stretched-out tracheoles (indicating parous mosquitoes). Non-parous mosquitos are considered young in this case, whereas parous mosquitos are considered old and may carry the malaria parasites, having had multiple blood-feedings [ 28 ]. Unfortunately, these dissections tend to be laborious and time-consuming, especially when dissecting large numbers of mosquitoes, and are impractical on a large scale. Furthermore, the reliability of mosquito dissections is limited by their reproductive history. For instance, a female mosquito can have more than one blood meal but still not oviposit, a scenario known as gonotrophic discordance or pre-gravid blood-meal [ 29 ]. Moreover, since the gonotrophic cycles of Anopheles mosquitoes can be as short as 2–3 days under optimal climatic conditions [ 30 , 31 ], it is possible for parous mosquitoes to be relatively young, and in rare cases, nulliparous mosquitoes to be several days old due to the scarcity of blood meals (e.g. when ITNs coverage and usage is high). Therefore, using parity alone to distinguish between epidemiologically distinct age categories of adult mosquitoes, especially in the context of malaria transmission, which requires 10–14 days of incubation [ 32 ], is not always realistic. All these concerns suggest the need for alternative age-grading techniques that are easy to perform cheaply at scale and can provide accurate representations of epidemiologically important mosquito age categories and populations. The alternative mosquito age-grading methods currently include the analysis of cuticular hydrocarbon patterns in a gas chromatograph [ 33 ] and gene transcription [ 34 – 36 ]. Near-infrared spectroscopy (NIRS) (12,500 cm − 1 to 4,000 cm − 1 frequencies) [ 37 ], which involves passing infrared light through a mosquito sample to measure absorbance or reflectance of the organic compound functional groups, has also been used to estimate ages for various mosquito species of both laboratory-reared and wild collected mosquitoes [ 38 – 44 ]. More recently, mid-infrared spectroscopy (MIRS) has been used to predict and estimate mosquito age, recording the biochemical composition of mosquito samples at longer wavelength frequencies [ 45 – 47 ]. In addition, machine learning (ML) techniques, including convolutional neural networks, have been utilized to differentiate MIRS spectra associated with distinct mosquito ages and species in both laboratory and wild mosquitoes [ 46 , 47 ]. The infrared based systems have so far been successful for various applications on three of the four main African malaria vectors (i.e. An. gambiae s.s, An. arabiensis and An. coluzzii [ 46 ], but have yet to be demonstrated for An. funestus . The goal of this study was therefore to test whether a similar ML-MIRS approach could classify adult female An. funestus mosquitoes derived from wild-caught larvae into two epidemiologically relevant age categories: young (0–9 days old, too young to have mature Plasmodium sporozoites in their salivary glands) and old (10 days or older, potentially carrying mature Plasmodium sporozoites given the right climatic conditions), factoring in a parasite incubation period of 10–14 days. Methods Mosquito collection Third and fourth instar mosquito larvae were collected from known aquatic habitats of An. funestus in five different villages in Southeastern Tanzania, namely Tulizamoyo (8.3669°S, 36.7336°E), Kilisa (8.3721°S, 36.5584°E), Lupiro (8.38333°S, 36.66667°E), Ikwambi (7.9833°S, 36.8184°E), and Ruaha (8.9068°S, 36.7185°E). The larvae were transported to the vector biology laboratory (VectorSphere), at Ifakara Health Institute for further rearing. The larvae were kept in water from their natural breeding habitats and were fed Tetramin® fish food. Once they pupated, the pupae were separated from the larvae and placed in emergence cages. The emergent adult mosquitoes were maintained at 26–28°C, 70–85% relative humidity and a 12:12 hour light/dark photoperiod, on a 10% sugar solution diet. Mosquito preservation and scanning The female adults were collected and individually preserved according to their age, from 1 to 16 days old. A total of 2084 mosquitoes were collected. The female mosquitoes were killed using chloroform and subsequently stored in separate 1.5 ml microcentrifuge tubes containing silica gel for desiccation. The heads and thoraces of the individual female mosquitoes were scanned using an Attenuated total reflection - Fourier transform infrared spectrometer (ATR – FT-IR) to obtain mid-infrared spectra with a resolution of 2 cm − 1 at 4000–400 cm − 1 frequencies as previously described, complete with background spectral calibration [ 45 , 48 , 49 ]. For each sample, 16 sample scans were averaged to obtain the primary output spectrum [ 46 ]. Mosquito identification Though the field collections had been done in known An. funestus habitats, it was necessary to confirm the identity of the mosquitoes and eliminate any unwanted species. This was accomplished primarily by morphology-based taxonomy using keys of Afro-tropical Anopheles [ 50 ] but was complemented by PCR identification to sort between sibling species in the An, funestus group. Wild An. funestus complex DNA was extracted from the two legs of adult female mosquitoes. The two legs of an individual An. funestus mosquito were placed separately in a 1.5 ml micro-centrifuge tube, followed by 20µl of TE buffer (Tris -EDTA), and incubated at 95 o C for 15 minutes. PCR was then used to differentiate An. funestus from other sibling species, using species-specific primers targeting the non-coding region of ITS2 using the protocol by Koekmoer et al. , [ 51 ]. The PCR reaction was performed in a 25µl volume, consisting of a PCR mixture of 2.5µl 10x reaction buffer, 25mM MgCl 2 , 10pmol/µl of each primer, 8mM of each dNTP, 5 units of thermo-stable Taq DNA polymerase, and 3µl of DNA template. The PCR products were analysed by electrophoresis in 2.5% agarose gel stained with classic view DNA dye for visualization of DNA bands. Only An. funestus mosquitoes were considered for further analysis, and any other species discarded. Machine learning Mosquito spectra with low intensity, abnormal background or atmospheric interferences (with water vapor and carbon dioxide) were discarded [ 45 ]. The data from the remaining spectra (N = 2084) were processed and analysed in Python 3.9, using Scikit-learn [ 52 ], and Tensorflow 2.0 [ 53 , 54 ]. The data were rescaled using the Standard Scaler algorithm, with a mean of 0 and a standard deviation of 1. Using the algorithm stratified shuffle split , the dataset was split into training (n = 1875) and test/unseen ( n = 209) sets. To train the supervised ML models, An. funestus ages were used as training labels. An. funestus , ranging from 1 to 16 days old, were divided into two epidemiologically relevant age categories taking into consideration the incubation period of malaria parasites of 10–14 [ 32 ]. The first group included An. funestus that were between 1 to 9 days old and were considered young and incapable of transmitting malaria (i.e., non-infectious age group). The second group included An. funestus that were between 10 to 16 days old and were considered old enough to be capable of transmitting malaria given the right environmental conditions (i.e., potentially infectious). Multiple standard machine learning (ML) classifiers, including k-nearest neighbours (KNN), logistic regression (LR), support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost), were compared to determine which model predicted the data with the highest classification accuracy. The best-performing model was further optimized by fine-tuning its hyperparameters. The top 100 spectral features (wavenumbers) with the most influence on the model predictions were identified and utilized to reduce the dimensionality of the spectra data, followed by retraining of the best ML classifier. Moreover, two Multi-layer Perceptron (MLP) models were trained by reducing the dimensionality of the spectra data using different inputs: 1) the top 100 features extracted from the best performing ML classifier, and 2) principal components using Scikit-learn library. Both MLP models had six fully connected layers, each containing 500 neurons, to enable the model to learn from the network's weights as previously demonstrated [ 55 ]. To prevent overfitting, a dropout layer with a rate of 0.5 was used, and early stopping was implemented when the validation loss could no longer improve after 400 iterations [ 56 , 57 ]. The model's performance was evaluated using K-fold cross-validation ( k = 5) to ensure an unbiased assessment of the standard ML and MLP models, as previously described [ 37 ]. To assess the ability of the optimized models to identify all positive instances and avoid false negatives, the recall score (i.e. sensitivity or true positive rate) was estimated as a ratio of correctly age-classified An. funestus to the total number of An. funestus in the respective age category in the dataset. Moreover, to measure the ability of the models to avoid false positives, the precision score (i.e. the positive predictive value) was estimated as a ratio of correctly age-classified An. funestus to the total number of predicted positive instances of the respective age categories. Lastly, we calculated the F1 score, which balances both precision and recall scores by giving equal weight to both measures. This score provides a single value that represents the overall performance of the model in terms of its ability to correctly classify positive and negative cases. A higher F1 score signifies a better model performance, where a maximum value of 1 represents flawless precision and recall. Results Predicting An. funestus age classes using standard machine learning models In the initial comparison of standard ML models, XGBoost emerged as the best classifier with the highest prediction accuracy and lowest standard deviation, achieving 84% accuracy (Fig. 1A). After optimizing the parameters, the XGBoost model was able to classify spectra that were previously unseen with an overall accuracy of 87%. It achieved accuracies of 89% and 84% for young (1–9 days old) and old (10–16 days old) An. funestus females respectively (Fig. 1B). The recall scores (i.e. sensitivity or true positive rates) of this model were 0.89 and 0.84 for the young and old mosquitoes respectively, while its precision scores (i.e. the positive predictive value) were 0.87 for both age categories (Table 1 ). Figure 1. Machine learning prediction of An. funestus age classes. A) Comparison of standard ML classifiers in predicting An. funestus age classes; KNN: K-nearest neighbours, LR: Logistic regression, SVM: Support vector machine, RF: Random Forest, XGBoost: Gradient boosting, MLP: Multilayer perceptron. B) Confusion matrix for predicting the age class of An. funestus using XGBoost on an unseen dataset, results for the ML trained with all spectra features. From the initial XGBoost model, we identified the spectral features that were most important for the prediction. This analysis aimed to reduce the number of training features and enhance the accuracy of the model during retraining (Fig. 2A). When the XGBoost classifier was retrained with the top 100 features, the classification accuracy increased to 89%, correctly predicting young and old An. funestus females with 92% and 85% accuracies respectively (Fig. 2B). Figure 2. A) Relative importance of XGBoost features that have the most influence in predicting the age classes of An. funestus . B) Confusion matrix for predicting the age class of An. funestus using XGBoost on an unseen dataset, the results for the ML retrained with important features/wavenumbers (n = 100) identified by the initial XGBoost model. Prediction of An. funestus age classes using MLP models We explored the possibility of improving the accuracy by training the MLP classifier using the important wavenumbers ( n = 100) identified in the XGBoost predictions. As a result, the MLP achieved an improved accuracy of 94.5% in the unseen test data (Fig. 3A), correctly distinguishing between young and old An. funestus females with accuracies of 95% and 94%, respectively (Fig. 3B). Lastly, in a previous study, we presented evidence that employing PCA with eight components effectively reduces the dimensionality of the spectra data [ 47 ]. This reduction in dimensionality not only preserved a substantial portion of the data variability, but also mitigated overfitting while enhancing the signal-to-noise ratio. By utilizing a reduced set of features, we trained the MLP model to improve its predictive performance [ 47 ]. In the present study, when PCA was utilized to reduce the dimensionality of the spectra data, the MLP classifier achieved an overall accuracy of 93% for both young and old An. funestus mosquitoes (Fig. 3C). Figure 3. A) MLP Training and validation accuracy for An. funestus age classes as training time increases (epoch). Confusion matrix for predicting the age class of An. funestus ; Panel B shows the results for the MLP trained with important features/wavenumbers (n = 100) identified by the XGBoost. Panel C shows the results for the MLP method trained with eight principal components. Table 1 Precision, recall, and f1 score of XGBoost and multi-layer perceptron (MLP) models for predicting age categories of An. funestus Model Age classes Precision Recall F1 -score No. test samples XGBoost 1 1–9 0.87 0.89 0.88 113 10–16 0.87 0.84 0.86 96 XGBoost 2 1–9 0.88 0.92 0.90 113 10–16 0.90 0.85 0.88 96 MLP 1 1–9 0.95 0.95 0.95 113 10–16 0.94 0.94 0.94 96 MLP 2 1–9 0.94 0.93 0.93 113 10–16 0.92 0.93 0.92 96 * XGBoost 1 : Trained with all MIRS wavenumbers ( n = 1665), XGBoost 2 : Trained with spectral features extracted based on feature importance summaries ( n = 100), MLP 1 : Trained with spectral features extracted based on feature importance summaries ( n = 100), MLP 2 : Trained with PCA as a dimensionality reduction technique. Discussion An. funestus mosquitoes are currently the major vector of malaria transmission in Tanzania, accounting for over 80% of malaria transmission [ 17 , 20 – 22 ]. An. funestus tends to have better survival rates [ 25 ], and is generally a slow-growing mosquito, which adds to the challenge of studying its demographic characteristics and how these might influence pathogen transmission. Here, we presented a rapid age-grading technique that has potential to replace the traditional methods like dissections, which are time-consuming and challenging to apply on a large scale. Using 2084 spectra data points, we trained machine learning models that classify the epidemiologically relevant age groups of An. funestus mosquitoes reared from wild larvae using water from the same habitats, but under laboratory conditions. The models correctly distinguished between the young An. funestus females (1–9 days old) and the older ones (10–16 days old) based on the MIR spectra indicative of the varying biochemical composition of the mosquito cuticles [ 58 ]. While this was the first demonstration of the effectiveness of this technique for predicting the age of An. funestus mosquitoes, the approach of combining infrared spectra and machine learning models has been widely demonstrated for predicting different indicators including age, blood meals, infection status, and insecticide resistance profiles of other Anopheles species [ 46 , 48 , 55 ]. If validated on field collected adults, these findings could be a step towards wider applications of this approach for malaria vector surveillance in settings with different vector species. In settings such as rural south-eastern Tanzania where An. funestus is the dominant malaria vector [ 17 , 20 ] it is particularly important that vector surveillance programs are expanded to include this vector species. Indeed, the successful demonstration that this technique on An. funestus , which is one of the most efficient and also most widespread malaria vectors in Africa [ 59 ], expands the rage of utility of this technique for a much broader application for malaria vector surveys in different parts of Africa. One of the key concerns regarding previous applications of MIRS-ML based approaches for entomological assessments is that, with exception of some cases [ 46 ], these methods have been rarely validated for wild-caught malaria vectors in field settings. Here, An. funestus mosquitoes were collected as larvae from various villages and breeding habitats, to account for genetic variation, variation in larval food sources and microbiome, and to maintain some characteristics of the natural ecosystems. The success of this analysis and the high accuracies obtained may therefore be indicative of the potential of the approach for predicting key mosquito attributes in field settings. However, it is unknown whether specific climatic factors could influence the prediction and generalizability of MIRS-ML approach. Future studies should therefore test the generalisability of this approach across different populations of wild mosquitoes. This study classified mosquitoes only as young (1–9 days old) or old (10–16 days old) and did not attempt to classify them at specific chronological ages because the sample size was not large enough to test it. However, the chosen age classes represent the typical epidemiological distinction relevant to the transmission of malaria parasites, which, under standard climatic conditions, requires that a vector must be at least 10 days old [ 32 ]. However, it may fail to capture variations in MIR spectra or the small biochemical changes that occur within a mosquito cuticle after each ageing day (such as chronological age from 1 up to 16) [ 45 ]. Moreover, it has been demonstrated that calibrating machine learning models based on physiological age (which considers key life cycle processes such as blood-feeding and oviposition) may be more useful than simply relying on chronological age classifications [ 38 , 60 ]. In our study, mosquitoes were all sugar fed, and therefore physiological age was not assessed. Future efforts should assess key differences in these approaches and evaluate models trained on biological age and chronological age to determine which ones are most practical and most generalizable. An obvious next step is therefore to investigate any correlations that might exist between the machine-classified age categories and the epidemiology of malaria in human populations. To improve the classification accuracy of our model, the XGBoost feature importance was relied upon to reduce the number of spectral features from 1665 to 100. This dimensionality reduction significantly lowered the noise and redundant features in the MIR spectra data. The important features were mostly associated with proteins, with the most influential peak (1700 cm − 1 ) being the band associated with the amide bond from proteins. The region around 3000 cm − 1 , which is also related to proteins, was also found to be important in the model prediction. This implies that the model is learning from protein-based biological traits that vary depending on the age of the mosquito [ 46 ]. Moreover, when PCA was used to reduce the dimensionality of the spectra from 1665 features to 8 principal components [ 47 ], the prediction accuracy matched that of the MLP model trained with the top 100 biological features as identified from the XGBoost model. This suggests that machine learning models may perform better when trained with fewer features that explain more variation in the data, rather than many redundant features that introduce noise into the model. Moreover, as observed previously, reducing the dimensionality of the spectra data reduces the computational resources needed to train machine learning models [ 47 ]. Future research should investigate the effects of rearing wild An. funestus larvae in the insectary on the predictive accuracies of MIRS-ML approach for mosquito age-classification as this could impact the generalizability of the findings. Conclusion This study demonstrates the classification of adult female An. funestus into distinct and epidemiologically relevant age categories using a MIRS-ML approach. In conjunction with prior research conducted on other Anopheles mosquitoes, this study suggests that the applicability of this approach can be extended to evaluate various entomological attributes in An. funestus . The MIRS-ML approach proves to be quick, cost-effective, and has the potential to significantly enhance An. funestus surveillance efforts, thereby contributing valuable insights to national malaria control programs, particularly in resource-constrained settings where this vector is highly prevalent. Nonetheless, further research is needed to validate the MIRS-ML approach in field conditions, using adult An. funestus populations and other vector species within malaria-endemic communities, and to examine how the machine-classified age categories correlate with the epidemiological strata of malaria in human populations. Abbreviations MIRS: Mid-Infrared spectroscopy; NIRS: Near-Infrared spectroscopy; PCR: Polymerase Chain Reaction; DL: Deep learning; ITNs: Insecticide-treated nets; MIRS-ML: Mid-infrared spectroscopy and machine learning. Declarations Acknowledgements The authors sincerely appreciate all field technicians who assisted in the collection of wild An. funestus larvae, as well as the rearing and handling of adult mosquitoes. The authors also express their gratitude to the administration team for their continuous administrative assistance. Additionally, we are grateful to the community members and local government officials in the districts of Ulanga and Kilombero for their unwavering support throughout this study. Authors’ contributions EPM, DJS, SB, FB, and FOO conceived the study. EPM, SA, FOO, and DJS developed the study's protocol. DJS collected the data, EPM carried out data analysis and ML training. EPM wrote the manuscript. EPM, DJS, SHM, IHM, MGJ, KW, SB, FB and FOO reviewed and edited drafts of the manuscript. All authors have read and approved the final manuscript. Funding This study was supported by a Howard Hughes Medical Institute (HHMI)-Gates International Research Scholarship (Grant No. OPP1099295) awarded to FOO and the Medical Research Council (MRC) [MR/P025501/1] awarded to FB. EPM was supported by the Wellcome Trust Masters Fellowship in Tropical Medicine & Hygiene (Grant No. 214643/Z/18/Z). FB is supported by the Academy Medical Sciences Springboard Award (ref:SBF007\100094). SAB is supported by the Bill and Melinda Gates Foundation (INV-030025) and Royal Society (ICA/R1/191238). Code, data, and materials availability The mid-infrared spectral datasets generated and analysed during the current study, as well as code for the analyses is available at [GitHub]. Ethics approval and consent to participate Ethical approval for this study was obtained from Ifakara Health Institute Institutional Review Board (Ref. IHI/IRB/EXT/No: 005-2018), and from the Medical Research Coordinating Committee (MRCC) at the National Institute of Medical Research (NIMR), Ref: NIMR/HQ/R.8c/Vol. II/880. Consent for publication All authors have read and approved the final manuscript. Permission to publish this work was obtained from National Institute of Medical Research (NIMR), Ref: NIMR/HQ/P.12 VOL.XXXVI/48. Competing interests The authors declare that they have no competing interests. References WHO. World malaria report. 2022. Winzeler EA, Manary MJ. Drug resistance genomics of the antimalarial drug artemisinin. Genome Biol [Internet]. 2014;15:544. Available from: https://doi.org/10.1186/s13059-014-0544-6 Mbengue A, Bhattacharjee S, Pandharkar T, Liu H, Estiu G, Stahelin R V, et al. A molecular mechanism of artemisinin resistance in Plasmodium falciparum malaria. Nature [Internet]. 2015;520:683–7. Available from: https://doi.org/10.1038/nature14412 Siddiqui G, Srivastava A, Russell AS, Creek DJ. Multi-omics Based Identification of Specific Biochemical Changes Associated With PfKelch13-Mutant Artemisinin-Resistant Plasmodium falciparum. J Infect Dis [Internet]. 2017;215:1435–44. Available from: https://doi.org/10.1093/infdis/jix156 Weill M, Lutfalla G, Mogensen K, Chandre F, Berthomieu A, Berticat C, et al. Insecticide resistance in mosquito vectors. Nature [Internet]. 2003;423:136–7. Available from: http://www.ncbi.nlm.nih.gov/entrez/query. fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=12736674 Sokhna C, Ndiath MO, Rogier C. The changes in mosquito vector behaviour and the emerging resistance to insecticides will challenge the decline of malaria. Clin Microbiol Infect [Internet]. 2013;19:902–7. Available from: https://www.sciencedirect.com/science/article/pii/S1198743X14629748 Bosco AB, Anderson K, Gresty K, Prosser C, Smith D, Nankabirwa JI, et al. Molecular surveillance reveals the presence of pfhrp2 and pfhrp3 gene deletions in Plasmodium falciparum parasite populations in Uganda, 2017–2019. Malar J [Internet]. 2020;19:300. Available from: https://doi.org/10.1186/s12936-020-03362-x Thomson R, Beshir KB, Cunningham J, Baiden F, Bharmal J, Bruxvoort KJ, et al. pfhrp2 and pfhrp3 Gene Deletions That Affect Malaria Rapid Diagnostic Tests for Plasmodium falciparum: Analysis of Archived Blood Samples From 3 African Countries. J Infect Dis [Internet]. 2019;220:1444–52. Available from: https://doi.org/10.1093/infdis/jiz335 Funwei R, Nderu D, Nguetse CN, Thomas BN, Falade CO, Velavan TP, et al. Molecular surveillance of pfhrp2 and pfhrp3 genes deletion in Plasmodium falciparum isolates and the implications for rapid diagnostic tests in Nigeria. Acta Trop [Internet]. 2019;196:121–5. Available from: https://www.sciencedirect.com/science/article/pii/S0001706X1831177X Berzosa P, González V, Taravillo L, Mayor A, Romay-Barja M, García L, et al. First evidence of the deletion in the pfhrp2 and pfhrp3 genes in Plasmodium falciparum from Equatorial Guinea. Malar J [Internet]. 2020;19:99. Available from: https://doi.org/10.1186/s12936-020-03178-9 Agaba BB, Yeka A, Nsobya S, Arinaitwe E, Nankabirwa J, Opigo J, et al. Systematic review of the status of pfhrp2 and pfhrp3 gene deletion, approaches and methods used for its estimation and reporting in Plasmodium falciparum populations in Africa: review of published studies 2010–2019. Malar J [Internet]. 2019;18:355. Available from: https://doi.org/10.1186/s12936-019-2987-4 WHO. World Malaria report 2021. 2021. Parpia AS, Ndeffo-Mbah ML, Wenzel NS, Galvani AP. Effects of response to 2014–2015 Ebola outbreak on deaths from malaria, HIV/AIDS, and tuberculosis, West Africa. Emerg Infect Dis. Centers for Disease Control and Prevention; 2016;22:433. Sherrard-Smith E, Hogan AB, Hamlet A, Watson OJ, Whittaker C, Winskill P, et al. The potential public health consequences of COVID-19 on malaria in Africa. Nat Med. Nature Publishing Group US New York; 2020;26:1411–6. Bhatt S, Weiss DJ, Cameron E, Bisanzio D, Mappin B, Dalrymple U, et al. The effect of malaria control on Plasmodium falciparum in Africa between 2000 and 2015. Nature. 2015; Lwetoijera DW, Harris C, Kiware SS, Dongus S, Devine GJ, McCall PJ, et al. Increasing role of Anopheles funestus and Anopheles arabiensis in malaria transmission in the Kilombero Valley, Tanzania. Malar J. 2014;13. Kaindoa EW, Matowo NS, Ngowo HS, Mkandawile G, Mmbando A, Finda M, et al. Interventions that effectively target Anopheles funestus mosquitoes could significantly improve control of persistent malaria transmission in south–eastern Tanzania. PLoS One [Internet]. Public Library of Science; 2017;12:e0177807. Available from: https://doi.org/10.1371/journal.pone.0177807 Djamouko-Djonkam L, Nkahe DL, Kopya E, Talipouo A, Ngadjeu CS, Doumbe-Belisse P, et al. Implication of Anopheles funestus in malaria transmission in the city of Yaoundé, Cameroon TT - Implication d’Anopheles funestus dans la transmission du paludisme dans la ville de Yaoundé au Cameroun. Parasite [Internet]. 2020/02/12. EDP Sciences; 2020;27:10. Available from: https://pubmed.ncbi.nlm.nih.gov/32048986 Ogola EO, Fillinger U, Ondiba IM, Villinger J, Masiga DK, Torto B, et al. Insights into malaria transmission among Anopheles funestus mosquitoes, Kenya. Parasit Vectors [Internet]. 2018;11:577. Available from: https://doi.org/10.1186/s13071-018-3171-3 Mapua SA, Hape EE, Kihonda J, Bwanary H, Kifungo K, Kilalangongono M, et al. Persistently high proportions of plasmodium-infected Anopheles funestus mosquitoes in two villages in the Kilombero valley, South-Eastern Tanzania. Parasite Epidemiol Control [Internet]. 2022;18:e00264. Available from: https://www.sciencedirect.com/science/article/pii/S2405673122000289 Swai JK, Mmbando AS, Ngowo HS, Odufuwa OG, Finda MF, Mponzi W, et al. Protecting migratory farmers in rural Tanzania using eave ribbons treated with the spatial mosquito repellent, transfluthrin. Malar J [Internet]. 2019;18:414. Available from: https://doi.org/10.1186/s12936-019-3048-8 Matowo NS, Kulkarni MA, Messenger LA, Jumanne M, Martin J, Mallya E, et al. Differential impact of dual-active ingredient long-lasting insecticidal nets on primary malaria vectors: a secondary analysis of a 3-year, single-blind, cluster-randomised controlled trial in rural Tanzania. Lancet Planet Heal [Internet]. Elsevier; 2023;7:e370–80. Available from: https://doi.org/10.1016/S2542-5196(23)00048-7 Takken W, Verhulst NO. Host Preferences of Blood-Feeding Mosquitoes. Annu Rev Entomol. 2013;58:433–53. Limwagu AJ, Kaindoa EW, Ngowo HS, Hape E, Finda M, Mkandawile G, et al. Using a miniaturized double-net trap (DN-Mini) to assess relationships between indoor–outdoor biting preferences and physiological ages of two malaria vectors, Anopheles arabiensis and Anopheles funestus. Malar J [Internet]. 2019;18:282. Available from: https://doi.org/10.1186/s12936-019-2913-9 Midega JT, Mbogo CM, Mwnambi H, Wilson MD, Ojwang G, Mwangangi JM, et al. Estimating dispersal and survival of Anopheles gambiae and Anopheles funestus along the Kenyan coast by using mark-release-recapture methods. J Med Entomol [Internet]. 2007;44:923–9. Available from: http://www.pubmedcentral.nih.gov/articlerender. fcgi?artid=2705338&tool=pmcentrez&rendertype=abstract Coetzee M, Koekemoer LL. Molecular systematics and insecticide resistance in the major African malaria vector Anopheles funestus. Annu Rev Entomol. Annual Reviews; 2013;58:393–412. Nambunga IH, Ngowo HS, Mapua SA, Hape EE, Msugupakulya BJ, Msaky DS, et al. Aquatic habitats of the malaria vector Anopheles funestus in rural south-eastern Tanzania. Malar J [Internet]. 2020;19:219. Available from: https://doi.org/10.1186/s12936-020-03295-5 Detinova TS. Age-grouping methods in Diptera of medical importance with special reference to some vectors of malaria. Monogr Ser World Health Organ. Geneva: World Health Organization; 1962; Rao V. On Gonotrophic Discordance among certain Indian Anopheles. Indian J Malariol. 1947;1:43–50. Derek Charlwood J, Nenhep S, Sovannaroth S, Morgan JC, Hemingway J, Chitnis N, et al. ‘Nature or nurture’: survival rate, oviposition interval, and possible gonotrophic discordance among South East Asian anophelines. Malar J [Internet]. 2016;15:356. Available from: https://doi.org/10.1186/s12936-016-1389-0 Charlwood JD, Tomás EVE, Andegiorgish AK, Mihreteab S, LeClair C. ‘We like it wet’: a comparison between dissection techniques for the assessment of parity in Anopheles arabiensis and determination of sac stage in mosquitoes alive or dead on collection. PeerJ. PeerJ Inc.; 2018;6:e5155. Ohm JR, Baldini F, Barreaux P, Lefevre T, Lynch PA, Suh E, et al. Rethinking the extrinsic incubation period of malaria parasites. Parasit Vectors [Internet]. 2018;11:178. Available from: https://doi.org/10.1186/s13071-018-2761-4 Chen CS, Mulla MS, March RB, Chaney JD. Cuticular hydrocarbon patterns in Culex quinquefasciatus as influenced by age, sex, and geography. Bull Soc Vector Ecol. 1990;15:129–39. Cook PE, Hugo LE, Iturbe-Ormaetxe I, Williams CR, Chenoweth SF, Ritchie SA, et al. Predicting the age of mosquitoes using transcriptional profiles. Nat Protoc [Internet]. 2007;2:2796–806. Available from: https://doi.org/10.1038/nprot.2007.396 Hugo LE, Cook PE, Johnson PH, Rapley LP, Kay BH, Ryan PA, et al. Field Validation of a Transcriptional Assay for the Prediction of Age of Uncaged Aedes aegypti Mosquitoes in Northern Australia. PLoS Negl Trop Dis [Internet]. Public Library of Science; 2010;4:e608. Available from: https://doi.org/10.1371/journal.pntd.0000608 Wang M-H, Marinotti O, James AA, Walker E, Githure J, Yan G. Genome-Wide Patterns of Gene Expression during Aging in the African Malaria Vector Anopheles gambiae. PLoS One [Internet]. Public Library of Science; 2010;5:e13359. Available from: https://doi.org/10.1371/journal.pone.0013359 Burns DA, Ciurczak EW. Handbook of near-infrared analysis. CRC press; 2008. Mayagaya VS, Michel K, Benedict MQ, Killeen GF, Wirtz RA, Ferguson HM, et al. Non-destructive determination of age and species of Anopheles gambiae sl using near-infrared spectroscopy. Am J Trop Med Hyg. 2009;81:622. Ong OTW, Kho EA, Esperança PM, Freebairn C, Dowell FE, Devine GJ, et al. Ability of near-infrared spectroscopy and chemometrics to predict the age of mosquitoes reared under different conditions. Parasit Vectors [Internet]. 2020;13:160. Available from: https://doi.org/10.1186/s13071-020-04031-3 Sikulu-Lord MT, Maia MF, Milali MP, Henry M, Mkandawile G, Kho EA, et al. Rapid and Non-destructive Detection and Identification of Two Strains of Wolbachia in Aedes aegypti by Near-Infrared Spectroscopy. PLoS Negl Trop Dis. 2016;10. Lambert B, Sikulu-Lord MT, Mayagaya VS, Devine G, Dowell F, Churcher TS. Monitoring the Age of Mosquito Populations Using Near-Infrared Spectroscopy. Sci Rep. 2018;8. Sikulu-Lord MT, Devine GJ, Hugo LE, Dowell FE. First report on the application of near-infrared spectroscopy to predict the age of Aedes albopictus Skuse. Sci Rep. 2018;8. Sikulu M, Killeen GF, Hugo LE, Ryan PA, Dowell KM, Wirtz RA, et al. Near-infrared spectroscopy as a complementary age grading and species identification tool for African malaria vectors. Parasit Vectors. 2010;3:49. Sikulu MT, Majambere S, Khatib BO, Ali AS, Hugo LE, Dowell FE. Using a Near-Infrared Spectrometer to Estimate the Age of Anopheles Mosquitoes Exposed to Pyrethroids. PLoS One [Internet]. Public Library of Science; 2014;9:e90657. Available from: https://doi.org/10.1371/journal.pone.0090657 Gonzalez-Jimenez M, Babayan SA, Khazaeli P, Doyle M, Walton F, Reedy E, et al. Prediction of malaria mosquito species and population age structure using mid-infrared spectroscopy and supervised machine learning. Wellcome Open Res. 2019;4:76. Siria DJ, Sanou R, Mitton J, Mwanga EP, Niang A, Sare I, et al. Rapid age-grading and species identification of natural mosquitoes for malaria surveillance. Nat Commun [Internet]. 2022;13:1501. Available from: https://doi.org/10.1038/s41467-022-28980-8 Mwanga EP, Siria DJ, Mitton J, Mshani IH, González-Jiménez M, Selvaraj P, et al. Using transfer learning and dimensionality reduction techniques to improve generalisability of machine-learning predictions of mosquito ages from mid-infrared spectra. BMC Bioinformatics [Internet]. 2023;24:11. Available from: https://doi.org/10.1186/s12859-022-05128-5 Mwanga EPP, Mapua SAA, Siria DJJ, Ngowo HSS, Nangacha F, Mgando J, et al. Using mid-infrared spectroscopy and supervised machine-learning to identify vertebrate blood meals in the malaria vector, Anopheles arabiensis. Malar J [Internet]. 2019;18:187. Available from: https://doi.org/10.1186/s12936-019-2822-y Mwanga EP, Minja EG, Mrimi E, Jiménez MG, Swai JK, Abbasi S, et al. Detection of malaria parasites in dried human blood spots using mid-infrared spectroscopy and logistic regression analysis. Malar J. 2019; Coetzee M. Key to the females of Afrotropical Anopheles mosquitoes (Diptera: Culicidae). Malar J [Internet]. 2020;19:70. Available from: https://doi.org/10.1186/s12936-020-3144-9 Koekemoer LL, Kamau L, Hunt RH, Coetzee M. A cocktail polymerase chain reaction assay to identify members of the Anopheles funestus (Diptera: Culicidae) group. Am J Trop Med Hyg [Internet]. 2002;66:804–11. Available from: http://www.ncbi.nlm.nih.gov/entrez/query. fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=12224596 Pedregosa F, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res. 2011; Abadi M, Barham P, Chen J, Chen Z, Davis A, Dean J, et al. TensorFlow: A system for large-scale machine learning. Proc 12th USENIX Symp Oper Syst Des Implementation, OSDI 2016. 2016. Chollet F. Keras: The Python Deep Learning library. KerasIo. 2015; Mwanga EP, Siria DJ, Mitton J, Mshani IH, Jimenez MG, Selvaraj P, et al. Using transfer learning and dimensionality reduction techniques to improve generalisability of machine-learning predictions of mosquito ages from mid-infrared spectra. bioRxiv [Internet]. 2022;2022.07.26.501594. Available from: http://biorxiv.org/content/early/2022/ 07/28/2022.07.26.501594.abstract Nitish S, Geoffrey H, Alex K, Ilya S, Ruslan S. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. J Mach Learn Res. 2014; Prechelt L. Early stopping - But when? Lect Notes Comput Sci (including Subser Lect Notes Artif Intell Lect Notes Bioinformatics). 2012; Suarez E, Nguyen HP, Ortiz IP, Lee KJ, Kim SB, Krzywinski J, et al. Matrix-assisted laser desorption/ionization-mass spectrometry of cuticular lipid profiles can differentiate sex, age, and mating status of Anopheles gambiae mosquitoes. Anal Chim Acta [Internet]. 2011;706:157–63. Available from: http://www.sciencedirect.com/science/article/pii/S0003267011011639 Dia I, Guelbeogo MW, Ayala D. Advances and Perspectives in the Study of the Malaria Mosquito Anopheles funestus. In: Manguin S, editor. Rijeka: IntechOpen; 2013. p. Ch. 7. Available from: https://doi.org/10.5772/55389 Ntamatungiro AJ, Mayagaya VS, Rieben S, Moore SJ, Dowell FE, Maia MF. The influence of physiological status on age prediction of Anopheles arabiensis using near infra-red spectroscopy. Parasites and Vectors [Internet]. 2013;6:298. Available from: http://parasitesandvectors.biomedcentral.com/articles/ 10.1186/1756-3305-6-298 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 18 Mar, 2024 Read the published version in Parasites & Vectors → Version 1 posted Editorial decision: Revision requested 18 Feb, 2024 Reviews received at journal 01 Feb, 2024 Reviewers agreed at journal 20 Jan, 2024 Reviewers agreed at journal 17 Jan, 2024 Reviewers agreed at journal 16 Jan, 2024 Reviewers invited by journal 14 Jan, 2024 Editor assigned by journal 04 Jan, 2024 Submission checks completed at journal 04 Jan, 2024 First submitted to journal 04 Jan, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3834184","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":265210566,"identity":"8179a62a-c5a8-4c50-8dbb-4776bccee21a","order_by":0,"name":"Emmanuel P. Mwanga","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwUlEQVRIiWNgGAWjYBADOTYJxgbStBjDtEgQrSWxAaqWsBZ+sTOGnwt+2aX3STc3MHzcw1AnT8h9krNzjKVn9iXntskcbGCc8YyBsJcMbudukObtYc5tk0hsYOY5wCDBTMhh9rdzN//m7alPZ4NpYSOkxUA6d5s0z4/DCXAtPIS0SNzO/2bN23DcEOSwgzMOSEjOIKSFf3Za8m2eP9Xy8jPSHz74cMCGn2CIgQFjG4Q+QEJM/iFW4SgYBaNgFIxIAADPwDjnP1LIqAAAAABJRU5ErkJggg==","orcid":"","institution":"Ifakara Health Institute","correspondingAuthor":true,"prefix":"","firstName":"Emmanuel","middleName":"P.","lastName":"Mwanga","suffix":""},{"id":265210567,"identity":"9ca2f92f-ad93-4c7f-9dbc-cf92d79836c1","order_by":1,"name":"Doreen J. Siria","email":"","orcid":"","institution":"Ifakara Health Institute","correspondingAuthor":false,"prefix":"","firstName":"Doreen","middleName":"J.","lastName":"Siria","suffix":""},{"id":265210568,"identity":"b6f893f3-18d1-4461-a13f-a42d6fbb7b4f","order_by":2,"name":"Issa H. Mshani","email":"","orcid":"","institution":"Ifakara Health Institute","correspondingAuthor":false,"prefix":"","firstName":"Issa","middleName":"H.","lastName":"Mshani","suffix":""},{"id":265210569,"identity":"8dd62cc0-6ab3-418d-9dce-90a4a7781601","order_by":3,"name":"Sophia H. Mwinyi","email":"","orcid":"","institution":"Ifakara Health Institute","correspondingAuthor":false,"prefix":"","firstName":"Sophia","middleName":"H.","lastName":"Mwinyi","suffix":""},{"id":265210570,"identity":"7d00c93f-9b33-4c27-833c-cc8c31753625","order_by":4,"name":"Said Abbas","email":"","orcid":"","institution":"Ifakara Health Institute","correspondingAuthor":false,"prefix":"","firstName":"Said","middleName":"","lastName":"Abbas","suffix":""},{"id":265210571,"identity":"9bcded16-3450-4a0a-9466-03eb9b4799f2","order_by":5,"name":"Mario Gonzalez Jimenez","email":"","orcid":"","institution":"University of Glasgow","correspondingAuthor":false,"prefix":"","firstName":"Mario","middleName":"Gonzalez","lastName":"Jimenez","suffix":""},{"id":265210572,"identity":"88d7bbd6-31ed-4b27-a685-8953dc3a9797","order_by":6,"name":"Klaas Wynne","email":"","orcid":"","institution":"University of Glasgow","correspondingAuthor":false,"prefix":"","firstName":"Klaas","middleName":"","lastName":"Wynne","suffix":""},{"id":265210573,"identity":"ddfba79c-f43b-428b-9fdb-6376aec58544","order_by":7,"name":"Francesco Baldini","email":"","orcid":"","institution":"University of Glasgow","correspondingAuthor":false,"prefix":"","firstName":"Francesco","middleName":"","lastName":"Baldini","suffix":""},{"id":265210574,"identity":"7e78bee0-f40f-4742-9e63-8867b92a84cb","order_by":8,"name":"Simon A. Babayan","email":"","orcid":"","institution":"University of Glasgow","correspondingAuthor":false,"prefix":"","firstName":"Simon","middleName":"A.","lastName":"Babayan","suffix":""},{"id":265210575,"identity":"6da0b733-174e-458d-bb70-e0209f49b703","order_by":9,"name":"Fredros O. Okumu","email":"","orcid":"","institution":"Ifakara Health Institute","correspondingAuthor":false,"prefix":"","firstName":"Fredros","middleName":"O.","lastName":"Okumu","suffix":""}],"badges":[],"createdAt":"2024-01-04 09:37:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3834184/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3834184/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s13071-024-06209-5","type":"published","date":"2024-03-18T15:01:03+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":49297962,"identity":"eb68567d-7148-4bdc-a4ca-aeb2eab22650","added_by":"auto","created_at":"2024-01-08 08:24:23","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":240870,"visible":true,"origin":"","legend":"\u003cp\u003eMachine learning prediction of An. funestus age classes. \u003cstrong\u003eA)\u003c/strong\u003e Comparison of standard ML classifiers in predicting An. funestus age classes; KNN: K-nearest neighbours, LR: Logistic regression, SVM: Support vector machine, RF: Random Forest, XGBoost: Gradient boosting, MLP: Multilayer perceptron. \u003cstrong\u003eB)\u003c/strong\u003e Confusion matrix for predicting the age class of \u003cem\u003eAn. funestus\u003c/em\u003e using XGBoost on an unseen dataset, results for the ML trained with all spectra features.\u003c/p\u003e","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3834184/v1/f559e04eba2f905b544fadf0.jpg"},{"id":49297964,"identity":"3c6294d2-5dc5-44ec-b3ac-e8d1adf68a8d","added_by":"auto","created_at":"2024-01-08 08:24:23","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":496231,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA) \u003c/strong\u003eRelative importance of XGBoost features that have the most influence in predicting the age classes of \u003cem\u003eAn. funestus\u003c/em\u003e. \u003cstrong\u003eB)\u003c/strong\u003e Confusion matrix for predicting the age class of \u003cem\u003eAn. funestus\u003c/em\u003e using XGBoost on an unseen dataset, the results for the ML retrained with important features/wavenumbers (n = 100) identified by the initial XGBoost model.\u003c/p\u003e","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3834184/v1/4856abda6779643fe4ed413b.jpg"},{"id":49298190,"identity":"b684a83c-9d10-48fe-b3c7-109abce9bbf1","added_by":"auto","created_at":"2024-01-08 08:32:23","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":337150,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e MLP Training and validation accuracy for \u003cem\u003eAn. funestus\u003c/em\u003e age classes as training time increases (epoch). Confusion matrix for predicting the age class of \u003cem\u003eAn. funestus\u003c/em\u003e; \u003cstrong\u003ePanel B\u003c/strong\u003e shows the results for the MLP trained with important features/wavenumbers (n = 100) identified by the XGBoost. \u003cstrong\u003ePanel C\u003c/strong\u003e shows the results for the MLP method trained with eight principal components.\u003c/p\u003e","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3834184/v1/02e7443b6caab4564777f369.jpg"},{"id":53403596,"identity":"12a1207f-24b8-4fe0-aba7-19d33090b560","added_by":"auto","created_at":"2024-03-25 15:13:08","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":447847,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3834184/v1/77c97a56-6937-4ddd-9243-ffd74f7c33d2.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Rapid classification of epidemiologically relevant age categories of the malaria vector, Anopheles funestus","fulltext":[{"header":"Background","content":"\u003cp\u003eDespite significant investments in malaria control and research, there were still an estimated 249\u0026nbsp;million malaria cases and 619,000 deaths in 2021 globally, nearly all of which in sub-Saharan Africa [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Other than the poor economic conditions and weak health systems, the continued high burden of malaria in Africa is attributable to key biological threats, notably malaria resistance to drugs [\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], vector resistance to insecticides [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], increasing occurrence of malaria parasites evading detection by rapid diagnostic tests [\u003cspan additionalcitationids=\"CR8 CR9 CR10\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e], and disruptions from major disease outbreaks such as Ebola and COVID-19 [\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Effective vector control, primarily with insecticide treated nets (ITNs) and indoor residual spraying (IRS), has been the most important component of malaria control in Africa [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. However, its continued effectiveness requires active innovation to address the current threats, and improved understanding of the major vector species in different settings.\u003c/p\u003e \u003cp\u003e \u003cem\u003eAnopheles funestus\u003c/em\u003e is one of the four main malaria vector species in sub-Saharan Africa, the others being \u003cem\u003eAn. gambiae, An. arabiensis\u003c/em\u003e and \u003cem\u003eAn. coluzzii\u003c/em\u003e, and also one of the most widespread [\u003cspan additionalcitationids=\"CR17 CR18\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. \u003cem\u003eAn. funestus\u003c/em\u003e is particularly important in East and Southern Africa, where it is becoming the dominant malaria vector. For example, in parts of Tanzania, \u003cem\u003eAn. funestus\u003c/em\u003e is reported to be responsible for 86\u0026ndash;97% of all new malaria infections [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Its dominance is due to multiple factors, including i) being highly anthropophilic, and thus preferring to bite humans over other vertebrates [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], ii) being highly endophilic, i.e. preferring to bite inside human dwellings than outside [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e], iii) having significantly higher survival rates than other species [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], iv) being resistant to commonly used insecticides [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] and v) preferentially breeding in perennial habitats with year-round productivity [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Given its importance and dominance in malaria transmission systems, vector surveillance programs in the respective countries should be designed with special attention to this vector species.\u003c/p\u003e \u003cp\u003eBesides evaluating biting densities and \u003cem\u003ePlasmodium\u003c/em\u003e infection rates, accurately determining the age and survival of \u003cem\u003eAn. funestus\u003c/em\u003e is crucial for monitoring transmission dynamics and assessing the effectiveness of vector control interventions such as ITNs and IRS. Dissection of mosquito ovaries is still the main entomological technique for estimating the age of vector populations [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The dissections are usually performed under light microscopes to assess the reproductive history, specifically the parity status, of the mosquitoes. This involves observing whether the ovaries contain coiled tracheolar skeins (indicating non-parous mosquitoes) or stretched-out tracheoles (indicating parous mosquitoes). Non-parous mosquitos are considered young in this case, whereas parous mosquitos are considered old and may carry the malaria parasites, having had multiple blood-feedings [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Unfortunately, these dissections tend to be laborious and time-consuming, especially when dissecting large numbers of mosquitoes, and are impractical on a large scale.\u003c/p\u003e \u003cp\u003eFurthermore, the reliability of mosquito dissections is limited by their reproductive history. For instance, a female mosquito can have more than one blood meal but still not oviposit, a scenario known as gonotrophic discordance or pre-gravid blood-meal [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Moreover, since the gonotrophic cycles of \u003cem\u003eAnopheles\u003c/em\u003e mosquitoes can be as short as 2\u0026ndash;3 days under optimal climatic conditions [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], it is possible for parous mosquitoes to be relatively young, and in rare cases, nulliparous mosquitoes to be several days old due to the scarcity of blood meals (e.g. when ITNs coverage and usage is high). Therefore, using parity alone to distinguish between epidemiologically distinct age categories of adult mosquitoes, especially in the context of malaria transmission, which requires 10\u0026ndash;14 days of incubation [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], is not always realistic.\u003c/p\u003e \u003cp\u003eAll these concerns suggest the need for alternative age-grading techniques that are easy to perform cheaply at scale and can provide accurate representations of epidemiologically important mosquito age categories and populations. The alternative mosquito age-grading methods currently include the analysis of cuticular hydrocarbon patterns in a gas chromatograph [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e] and gene transcription [\u003cspan additionalcitationids=\"CR35\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. Near-infrared spectroscopy (NIRS) (12,500 cm\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e to 4,000 cm\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e frequencies) [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e], which involves passing infrared light through a mosquito sample to measure absorbance or reflectance of the organic compound functional groups, has also been used to estimate ages for various mosquito species of both laboratory-reared and wild collected mosquitoes [\u003cspan additionalcitationids=\"CR39 CR40 CR41 CR42 CR43\" citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eMore recently, mid-infrared spectroscopy (MIRS) has been used to predict and estimate mosquito age, recording the biochemical composition of mosquito samples at longer wavelength frequencies [\u003cspan additionalcitationids=\"CR46\" citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. In addition, machine learning (ML) techniques, including convolutional neural networks, have been utilized to differentiate MIRS spectra associated with distinct mosquito ages and species in both laboratory and wild mosquitoes [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. The infrared based systems have so far been successful for various applications on three of the four main African malaria vectors (i.e. \u003cem\u003eAn. gambiae s.s, An. arabiensis\u003c/em\u003e and \u003cem\u003eAn. coluzzii\u003c/em\u003e [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e], but have yet to be demonstrated for \u003cem\u003eAn. funestus\u003c/em\u003e. The goal of this study was therefore to test whether a similar ML-MIRS approach could classify adult female \u003cem\u003eAn. funestus\u003c/em\u003e mosquitoes derived from wild-caught larvae into two epidemiologically relevant age categories: young (0\u0026ndash;9 days old, too young to have mature \u003cem\u003ePlasmodium\u003c/em\u003e sporozoites in their salivary glands) and old (10 days or older, potentially carrying mature \u003cem\u003ePlasmodium\u003c/em\u003e sporozoites given the right climatic conditions), factoring in a parasite incubation period of 10\u0026ndash;14 days.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eMosquito collection\u003c/h2\u003e \u003cp\u003eThird and fourth instar mosquito larvae were collected from known aquatic habitats of \u003cem\u003eAn. funestus\u003c/em\u003e in five different villages in Southeastern Tanzania, namely Tulizamoyo (8.3669\u0026deg;S, 36.7336\u0026deg;E), Kilisa (8.3721\u0026deg;S, 36.5584\u0026deg;E), Lupiro (8.38333\u0026deg;S, 36.66667\u0026deg;E), Ikwambi (7.9833\u0026deg;S, 36.8184\u0026deg;E), and Ruaha (8.9068\u0026deg;S, 36.7185\u0026deg;E). The larvae were transported to the vector biology laboratory (VectorSphere), at Ifakara Health Institute for further rearing. The larvae were kept in water from their natural breeding habitats and were fed Tetramin\u0026reg; fish food.\u003c/p\u003e \u003cp\u003eOnce they pupated, the pupae were separated from the larvae and placed in emergence cages. The emergent adult mosquitoes were maintained at 26\u0026ndash;28\u0026deg;C, 70\u0026ndash;85% relative humidity and a 12:12 hour light/dark photoperiod, on a 10% sugar solution diet.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eMosquito preservation and scanning\u003c/h2\u003e \u003cp\u003eThe female adults were collected and individually preserved according to their age, from 1 to 16 days old. A total of 2084 mosquitoes were collected. The female mosquitoes were killed using chloroform and subsequently stored in separate 1.5 ml microcentrifuge tubes containing silica gel for desiccation. The heads and thoraces of the individual female mosquitoes were scanned using an Attenuated total reflection - Fourier transform infrared spectrometer (ATR \u0026ndash; FT-IR) to obtain mid-infrared spectra with a resolution of 2 cm\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e at 4000\u0026ndash;400 cm\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e frequencies as previously described, complete with background spectral calibration [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. For each sample, 16 sample scans were averaged to obtain the primary output spectrum [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eMosquito identification\u003c/h2\u003e \u003cp\u003eThough the field collections had been done in known \u003cem\u003eAn. funestus\u003c/em\u003e habitats, it was necessary to confirm the identity of the mosquitoes and eliminate any unwanted species. This was accomplished primarily by morphology-based taxonomy using keys of Afro-tropical \u003cem\u003eAnopheles\u003c/em\u003e [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e] but was complemented by PCR identification to sort between sibling species in the \u003cem\u003eAn, funestus\u003c/em\u003e group. Wild \u003cem\u003eAn. funestus\u003c/em\u003e complex DNA was extracted from the two legs of adult female mosquitoes. The two legs of an individual \u003cem\u003eAn. funestus\u003c/em\u003e mosquito were placed separately in a 1.5 ml micro-centrifuge tube, followed by 20\u0026micro;l of TE buffer (Tris -EDTA), and incubated at 95\u003csup\u003eo\u003c/sup\u003eC for 15 minutes. PCR was then used to differentiate \u003cem\u003eAn. funestus\u003c/em\u003e from other sibling species, using species-specific primers targeting the non-coding region of ITS2 using the protocol by Koekmoer \u003cem\u003eet al.\u003c/em\u003e, [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e]. The PCR reaction was performed in a 25\u0026micro;l volume, consisting of a PCR mixture of 2.5\u0026micro;l 10x reaction buffer, 25mM MgCl\u003csub\u003e2\u003c/sub\u003e, 10pmol/\u0026micro;l of each primer, 8mM of each dNTP, 5 units of thermo-stable Taq DNA polymerase, and 3\u0026micro;l of DNA template. The PCR products were analysed by electrophoresis in 2.5% agarose gel stained with classic view DNA dye for visualization of DNA bands. Only \u003cem\u003eAn. funestus\u003c/em\u003e mosquitoes were considered for further analysis, and any other species discarded.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eMachine learning\u003c/h2\u003e \u003cp\u003eMosquito spectra with low intensity, abnormal background or atmospheric interferences (with water vapor and carbon dioxide) were discarded [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. The data from the remaining spectra (N\u0026thinsp;=\u0026thinsp;2084) were processed and analysed in Python 3.9, using Scikit-learn [\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e], and Tensorflow 2.0 [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e, \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. The data were rescaled using the \u003cem\u003eStandard Scaler\u003c/em\u003e algorithm, with a mean of 0 and a standard deviation of 1.\u003c/p\u003e \u003cp\u003eUsing the algorithm \u003cem\u003estratified shuffle split\u003c/em\u003e, the dataset was split into training \u003cem\u003e(n\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1875) and test/unseen (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;209) sets. To train the supervised ML models, \u003cem\u003eAn. funestus\u003c/em\u003e ages were used as training labels. \u003cem\u003eAn. funestus\u003c/em\u003e, ranging from 1 to 16 days old, were divided into two epidemiologically relevant age categories taking into consideration the incubation period of malaria parasites of 10\u0026ndash;14 [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. The first group included \u003cem\u003eAn. funestus\u003c/em\u003e that were between 1 to 9 days old and were considered young and incapable of transmitting malaria (i.e., non-infectious age group). The second group included \u003cem\u003eAn. funestus\u003c/em\u003e that were between 10 to 16 days old and were considered old enough to be capable of transmitting malaria given the right environmental conditions (i.e., potentially infectious).\u003c/p\u003e \u003cp\u003eMultiple standard machine learning (ML) classifiers, including k-nearest neighbours (KNN), logistic regression (LR), support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost), were compared to determine which model predicted the data with the highest classification accuracy. The best-performing model was further optimized by fine-tuning its hyperparameters. The top 100 spectral features (wavenumbers) with the most influence on the model predictions were identified and utilized to reduce the dimensionality of the spectra data, followed by retraining of the best ML classifier.\u003c/p\u003e \u003cp\u003eMoreover, two Multi-layer Perceptron (MLP) models were trained by reducing the dimensionality of the spectra data using different inputs: 1) the top 100 features extracted from the best performing ML classifier, and 2) principal components using Scikit-learn library. Both MLP models had six fully connected layers, each containing 500 neurons, to enable the model to learn from the network's weights as previously demonstrated [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. To prevent overfitting, a dropout layer with a rate of 0.5 was used, and early stopping was implemented when the validation loss could no longer improve after 400 iterations [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]. The model's performance was evaluated using K-fold cross-validation (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;5) to ensure an unbiased assessment of the standard ML and MLP models, as previously described [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTo assess the ability of the optimized models to identify all positive instances and avoid false negatives, the recall score (i.e. sensitivity or true positive rate) was estimated as a ratio of correctly age-classified \u003cem\u003eAn. funestus\u003c/em\u003e to the total number of \u003cem\u003eAn. funestus\u003c/em\u003e in the respective age category in the dataset. Moreover, to measure the ability of the models to avoid false positives, the precision score (i.e. the positive predictive value) was estimated as a ratio of correctly age-classified \u003cem\u003eAn. funestus\u003c/em\u003e to the total number of predicted positive instances of the respective age categories. Lastly, we calculated the \u003cem\u003eF1\u003c/em\u003e score, which balances both precision and recall scores by giving equal weight to both measures. This score provides a single value that represents the overall performance of the model in terms of its ability to correctly classify positive and negative cases. A higher \u003cem\u003eF1\u003c/em\u003e score signifies a better model performance, where a maximum value of 1 represents flawless precision and recall.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e \u003cb\u003ePredicting\u003c/b\u003e \u003cb\u003eAn. funestus\u003c/b\u003e \u003cb\u003eage classes using standard machine learning models\u003c/b\u003e\u003c/p\u003e \u003cp\u003eIn the initial comparison of standard ML models, XGBoost emerged as the best classifier with the highest prediction accuracy and lowest standard deviation, achieving 84% accuracy (Fig.\u0026nbsp;1A). After optimizing the parameters, the XGBoost model was able to classify spectra that were previously unseen with an overall accuracy of 87%. It achieved accuracies of 89% and 84% for young (1\u0026ndash;9 days old) and old (10\u0026ndash;16 days old) \u003cem\u003eAn. funestus\u003c/em\u003e females respectively (Fig.\u0026nbsp;1B). The recall scores (i.e. sensitivity or true positive rates) of this model were 0.89 and 0.84 for the young and old mosquitoes respectively, while its precision scores (i.e. the positive predictive value) were 0.87 for both age categories (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cb\u003eFigure 1.\u003c/b\u003e Machine learning prediction of An. funestus age classes. \u003cb\u003eA)\u003c/b\u003e Comparison of standard ML classifiers in predicting An. funestus age classes; KNN: K-nearest neighbours, LR: Logistic regression, SVM: Support vector machine, RF: Random Forest, XGBoost: Gradient boosting, MLP: Multilayer perceptron. \u003cb\u003eB)\u003c/b\u003e Confusion matrix for predicting the age class of \u003cem\u003eAn. funestus\u003c/em\u003e using XGBoost on an unseen dataset, results for the ML trained with all spectra features.\u003c/p\u003e \u003cp\u003eFrom the initial XGBoost model, we identified the spectral features that were most important for the prediction. This analysis aimed to reduce the number of training features and enhance the accuracy of the model during retraining (Fig.\u0026nbsp;2A). When the XGBoost classifier was retrained with the top 100 features, the classification accuracy increased to 89%, correctly predicting young and old \u003cem\u003eAn. funestus\u003c/em\u003e females with 92% and 85% accuracies respectively (Fig.\u0026nbsp;2B).\u003c/p\u003e \u003cp\u003e \u003cb\u003eFigure 2. A)\u003c/b\u003e Relative importance of XGBoost features that have the most influence in predicting the age classes of \u003cem\u003eAn. funestus\u003c/em\u003e. \u003cb\u003eB)\u003c/b\u003e Confusion matrix for predicting the age class of \u003cem\u003eAn. funestus\u003c/em\u003e using XGBoost on an unseen dataset, the results for the ML retrained with important features/wavenumbers (n\u0026thinsp;=\u0026thinsp;100) identified by the initial XGBoost model.\u003c/p\u003e \u003cp\u003e \u003cb\u003ePrediction of\u003c/b\u003e \u003cb\u003eAn. funestus\u003c/b\u003e \u003cb\u003eage classes using MLP models\u003c/b\u003e\u003c/p\u003e \u003cp\u003eWe explored the possibility of improving the accuracy by training the MLP classifier using the important wavenumbers (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;100) identified in the XGBoost predictions. As a result, the MLP achieved an improved accuracy of 94.5% in the unseen test data (Fig.\u0026nbsp;3A), correctly distinguishing between young and old \u003cem\u003eAn. funestus\u003c/em\u003e females with accuracies of 95% and 94%, respectively (Fig.\u0026nbsp;3B).\u003c/p\u003e \u003cp\u003eLastly, in a previous study, we presented evidence that employing PCA with eight components effectively reduces the dimensionality of the spectra data [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. This reduction in dimensionality not only preserved a substantial portion of the data variability, but also mitigated overfitting while enhancing the signal-to-noise ratio. By utilizing a reduced set of features, we trained the MLP model to improve its predictive performance [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. In the present study, when PCA was utilized to reduce the dimensionality of the spectra data, the MLP classifier achieved an overall accuracy of 93% for both young and old \u003cem\u003eAn. funestus\u003c/em\u003e mosquitoes (Fig.\u0026nbsp;3C).\u003c/p\u003e \u003cp\u003e \u003cb\u003eFigure 3. A)\u003c/b\u003e MLP Training and validation accuracy for \u003cem\u003eAn. funestus\u003c/em\u003e age classes as training time increases (epoch). Confusion matrix for predicting the age class of \u003cem\u003eAn. funestus\u003c/em\u003e; \u003cb\u003ePanel B\u003c/b\u003e shows the results for the MLP trained with important features/wavenumbers (n\u0026thinsp;=\u0026thinsp;100) identified by the XGBoost. \u003cb\u003ePanel C\u003c/b\u003e shows the results for the MLP method trained with eight principal components.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePrecision, recall, and \u003cem\u003ef1\u003c/em\u003e score of XGBoost and multi-layer perceptron (MLP) models for predicting age categories of \u003cem\u003eAn. funestus\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAge classes\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eF1\u003c/em\u003e-score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNo. test samples\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eXGBoost 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u0026ndash;9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e113\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10\u0026ndash;16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e96\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eXGBoost 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u0026ndash;9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e113\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10\u0026ndash;16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e96\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eMLP 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u0026ndash;9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e113\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10\u0026ndash;16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e96\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eMLP 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u0026ndash;9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e113\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10\u0026ndash;16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e96\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e* \u003cb\u003eXGBoost 1\u003c/b\u003e: Trained with all MIRS wavenumbers (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1665), \u003cb\u003eXGBoost 2\u003c/b\u003e: Trained with spectral features extracted based on feature importance summaries (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;100), \u003cb\u003eMLP 1\u003c/b\u003e: Trained with spectral features extracted based on feature importance summaries (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;100), \u003cb\u003eMLP 2\u003c/b\u003e: Trained with PCA as a dimensionality reduction technique.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003e \u003cem\u003eAn. funestus\u003c/em\u003e mosquitoes are currently the major vector of malaria transmission in Tanzania, accounting for over 80% of malaria transmission [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. \u003cem\u003eAn. funestus\u003c/em\u003e tends to have better survival rates [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], and is generally a slow-growing mosquito, which adds to the challenge of studying its demographic characteristics and how these might influence pathogen transmission. Here, we presented a rapid age-grading technique that has potential to replace the traditional methods like dissections, which are time-consuming and challenging to apply on a large scale. Using 2084 spectra data points, we trained machine learning models that classify the epidemiologically relevant age groups of \u003cem\u003eAn. funestus\u003c/em\u003e mosquitoes reared from wild larvae using water from the same habitats, but under laboratory conditions. The models correctly distinguished between the young \u003cem\u003eAn. funestus\u003c/em\u003e females (1\u0026ndash;9 days old) and the older ones (10\u0026ndash;16 days old) based on the MIR spectra indicative of the varying biochemical composition of the mosquito cuticles [\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e]. While this was the first demonstration of the effectiveness of this technique for predicting the age of \u003cem\u003eAn. funestus\u003c/em\u003e mosquitoes, the approach of combining infrared spectra and machine learning models has been widely demonstrated for predicting different indicators including age, blood meals, infection status, and insecticide resistance profiles of other \u003cem\u003eAnopheles\u003c/em\u003e species [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. If validated on field collected adults, these findings could be a step towards wider applications of this approach for malaria vector surveillance in settings with different vector species.\u003c/p\u003e \u003cp\u003eIn settings such as rural south-eastern Tanzania where \u003cem\u003eAn. funestus\u003c/em\u003e is the dominant malaria vector [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] it is particularly important that vector surveillance programs are expanded to include this vector species. Indeed, the successful demonstration that this technique on \u003cem\u003eAn. funestus\u003c/em\u003e, which is one of the most efficient and also most widespread malaria vectors in Africa [\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e], expands the rage of utility of this technique for a much broader application for malaria vector surveys in different parts of Africa.\u003c/p\u003e \u003cp\u003eOne of the key concerns regarding previous applications of MIRS-ML based approaches for entomological assessments is that, with exception of some cases [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e], these methods have been rarely validated for wild-caught malaria vectors in field settings. Here, \u003cem\u003eAn. funestus\u003c/em\u003e mosquitoes were collected as larvae from various villages and breeding habitats, to account for genetic variation, variation in larval food sources and microbiome, and to maintain some characteristics of the natural ecosystems. The success of this analysis and the high accuracies obtained may therefore be indicative of the potential of the approach for predicting key mosquito attributes in field settings. However, it is unknown whether specific climatic factors could influence the prediction and generalizability of MIRS-ML approach. Future studies should therefore test the generalisability of this approach across different populations of wild mosquitoes.\u003c/p\u003e \u003cp\u003eThis study classified mosquitoes only as young (1\u0026ndash;9 days old) or old (10\u0026ndash;16 days old) and did not attempt to classify them at specific chronological ages because the sample size was not large enough to test it. However, the chosen age classes represent the typical epidemiological distinction relevant to the transmission of malaria parasites, which, under standard climatic conditions, requires that a vector must be at least 10 days old [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. However, it may fail to capture variations in MIR spectra or the small biochemical changes that occur within a mosquito cuticle after each ageing day (such as chronological age from 1 up to 16) [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. Moreover, it has been demonstrated that calibrating machine learning models based on physiological age (which considers key life cycle processes such as blood-feeding and oviposition) may be more useful than simply relying on chronological age classifications [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e]. In our study, mosquitoes were all sugar fed, and therefore physiological age was not assessed. Future efforts should assess key differences in these approaches and evaluate models trained on biological age and chronological age to determine which ones are most practical and most generalizable. An obvious next step is therefore to investigate any correlations that might exist between the machine-classified age categories and the epidemiology of malaria in human populations.\u003c/p\u003e \u003cp\u003eTo improve the classification accuracy of our model, the XGBoost feature importance was relied upon to reduce the number of spectral features from 1665 to 100. This dimensionality reduction significantly lowered the noise and redundant features in the MIR spectra data. The important features were mostly associated with proteins, with the most influential peak (1700 cm\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e) being the band associated with the amide bond from proteins. The region around 3000 cm\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e, which is also related to proteins, was also found to be important in the model prediction. This implies that the model is learning from protein-based biological traits that vary depending on the age of the mosquito [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. Moreover, when PCA was used to reduce the dimensionality of the spectra from 1665 features to 8 principal components [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e], the prediction accuracy matched that of the MLP model trained with the top 100 biological features as identified from the XGBoost model. This suggests that machine learning models may perform better when trained with fewer features that explain more variation in the data, rather than many redundant features that introduce noise into the model. Moreover, as observed previously, reducing the dimensionality of the spectra data reduces the computational resources needed to train machine learning models [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFuture research should investigate the effects of rearing wild \u003cem\u003eAn. funestus\u003c/em\u003e larvae in the insectary on the predictive accuracies of MIRS-ML approach for mosquito age-classification as this could impact the generalizability of the findings.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study demonstrates the classification of adult female \u003cem\u003eAn. funestus\u003c/em\u003e into distinct and epidemiologically relevant age categories using a MIRS-ML approach. In conjunction with prior research conducted on other \u003cem\u003eAnopheles\u003c/em\u003e mosquitoes, this study suggests that the applicability of this approach can be extended to evaluate various entomological attributes in \u003cem\u003eAn. funestus\u003c/em\u003e. The MIRS-ML approach proves to be quick, cost-effective, and has the potential to significantly enhance \u003cem\u003eAn. funestus\u003c/em\u003e surveillance efforts, thereby contributing valuable insights to national malaria control programs, particularly in resource-constrained settings where this vector is highly prevalent. Nonetheless, further research is needed to validate the MIRS-ML approach in field conditions, using adult \u003cem\u003eAn. funestus\u003c/em\u003e populations and other vector species within malaria-endemic communities, and to examine how the machine-classified age categories correlate with the epidemiological strata of malaria in human populations.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eMIRS: Mid-Infrared spectroscopy; NIRS: Near-Infrared spectroscopy; PCR: Polymerase Chain Reaction; DL: Deep learning; ITNs: Insecticide-treated nets; MIRS-ML: Mid-infrared spectroscopy and machine learning.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eThe authors sincerely appreciate all field technicians who assisted in the collection of wild \u003cem\u003eAn. funestus\u003c/em\u003e larvae, as well as the rearing and handling of adult mosquitoes. The authors also express their gratitude to the administration team for their continuous administrative assistance. Additionally, we are grateful to the community members and local government officials in the districts of Ulanga and Kilombero for their unwavering support throughout this study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAuthors\u0026rsquo; contributions\u003c/p\u003e\n\u003cp\u003eEPM, DJS, SB, FB, and FOO conceived the study. EPM, SA, FOO, and DJS developed the study\u0026apos;s protocol. DJS collected the data, EPM carried out data analysis and ML training. EPM wrote the manuscript. EPM, DJS, SHM, IHM, MGJ, KW, SB, FB and FOO reviewed and edited drafts of the manuscript. All authors have read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eThis study was supported by a Howard Hughes Medical Institute (HHMI)-Gates International Research Scholarship (Grant No. OPP1099295) awarded to FOO and the\u0026nbsp;Medical Research Council (MRC) [MR/P025501/1] awarded to FB. EPM was supported by the Wellcome Trust Masters Fellowship in Tropical Medicine \u0026amp; Hygiene (Grant No. 214643/Z/18/Z). FB is supported by the Academy Medical Sciences Springboard Award (ref:SBF007\\100094). SAB is supported by the Bill and Melinda Gates Foundation (INV-030025) and Royal Society (ICA/R1/191238).\u003c/p\u003e\n\u003cp\u003eCode, data, and materials availability\u003c/p\u003e\n\u003cp\u003eThe mid-infrared spectral datasets generated and analysed during the current study, as well as code for the analyses is available at [GitHub].\u003c/p\u003e\n\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eEthical approval for this study was obtained from Ifakara Health Institute Institutional Review Board (Ref. IHI/IRB/EXT/No: 005-2018), and from the Medical Research Coordinating Committee (MRCC) at the National Institute of Medical Research (NIMR), Ref: NIMR/HQ/R.8c/Vol. II/880.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConsent for publication\u003c/p\u003e\n\u003cp\u003eAll authors have read and approved the final manuscript. Permission to publish this work was obtained from National Institute of Medical Research (NIMR), Ref: NIMR/HQ/P.12 VOL.XXXVI/48. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eWHO. World malaria report. 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWinzeler EA, Manary MJ. Drug resistance genomics of the antimalarial drug artemisinin. Genome Biol [Internet]. 2014;15:544. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13059-014-0544-6\u003c/span\u003e\u003cspan address=\"10.1186/s13059-014-0544-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMbengue A, Bhattacharjee S, Pandharkar T, Liu H, Estiu G, Stahelin R V, et al. A molecular mechanism of artemisinin resistance in Plasmodium falciparum malaria. Nature [Internet]. 2015;520:683\u0026ndash;7. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nature14412\u003c/span\u003e\u003cspan address=\"10.1038/nature14412\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSiddiqui G, Srivastava A, Russell AS, Creek DJ. Multi-omics Based Identification of Specific Biochemical Changes Associated With PfKelch13-Mutant Artemisinin-Resistant Plasmodium falciparum. J Infect Dis [Internet]. 2017;215:1435\u0026ndash;44. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/infdis/jix156\u003c/span\u003e\u003cspan address=\"10.1093/infdis/jix156\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeill M, Lutfalla G, Mogensen K, Chandre F, Berthomieu A, Berticat C, et al. Insecticide resistance in mosquito vectors. Nature [Internet]. 2003;423:136\u0026ndash;7. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.ncbi.nlm.nih.gov/entrez/query.\u003c/span\u003e\u003cspan address=\"http://www.ncbi.nlm.nih.gov/entrez/query.\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003efcgi?cmd=Retrieve\u0026amp;db=PubMed\u0026amp;dopt=Citation\u0026amp;list_uids=12736674\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSokhna C, Ndiath MO, Rogier C. The changes in mosquito vector behaviour and the emerging resistance to insecticides will challenge the decline of malaria. Clin Microbiol Infect [Internet]. 2013;19:902\u0026ndash;7. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/science/article/pii/S1198743X14629748\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/science/article/pii/S1198743X14629748\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBosco AB, Anderson K, Gresty K, Prosser C, Smith D, Nankabirwa JI, et al. Molecular surveillance reveals the presence of pfhrp2 and pfhrp3 gene deletions in Plasmodium falciparum parasite populations in Uganda, 2017\u0026ndash;2019. Malar J [Internet]. 2020;19:300. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-020-03362-x\u003c/span\u003e\u003cspan address=\"10.1186/s12936-020-03362-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThomson R, Beshir KB, Cunningham J, Baiden F, Bharmal J, Bruxvoort KJ, et al. pfhrp2 and pfhrp3 Gene Deletions That Affect Malaria Rapid Diagnostic Tests for Plasmodium falciparum: Analysis of Archived Blood Samples From 3 African Countries. J Infect Dis [Internet]. 2019;220:1444\u0026ndash;52. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/infdis/jiz335\u003c/span\u003e\u003cspan address=\"10.1093/infdis/jiz335\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFunwei R, Nderu D, Nguetse CN, Thomas BN, Falade CO, Velavan TP, et al. Molecular surveillance of pfhrp2 and pfhrp3 genes deletion in Plasmodium falciparum isolates and the implications for rapid diagnostic tests in Nigeria. Acta Trop [Internet]. 2019;196:121\u0026ndash;5. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/science/article/pii/S0001706X1831177X\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/science/article/pii/S0001706X1831177X\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerzosa P, Gonz\u0026aacute;lez V, Taravillo L, Mayor A, Romay-Barja M, Garc\u0026iacute;a L, et al. First evidence of the deletion in the pfhrp2 and pfhrp3 genes in Plasmodium falciparum from Equatorial Guinea. Malar J [Internet]. 2020;19:99. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-020-03178-9\u003c/span\u003e\u003cspan address=\"10.1186/s12936-020-03178-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAgaba BB, Yeka A, Nsobya S, Arinaitwe E, Nankabirwa J, Opigo J, et al. Systematic review of the status of pfhrp2 and pfhrp3 gene deletion, approaches and methods used for its estimation and reporting in Plasmodium falciparum populations in Africa: review of published studies 2010\u0026ndash;2019. Malar J [Internet]. 2019;18:355. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-019-2987-4\u003c/span\u003e\u003cspan address=\"10.1186/s12936-019-2987-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWHO. World Malaria report 2021. 2021.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eParpia AS, Ndeffo-Mbah ML, Wenzel NS, Galvani AP. Effects of response to 2014\u0026ndash;2015 Ebola outbreak on deaths from malaria, HIV/AIDS, and tuberculosis, West Africa. Emerg Infect Dis. Centers for Disease Control and Prevention; 2016;22:433.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSherrard-Smith E, Hogan AB, Hamlet A, Watson OJ, Whittaker C, Winskill P, et al. The potential public health consequences of COVID-19 on malaria in Africa. Nat Med. Nature Publishing Group US New York; 2020;26:1411\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBhatt S, Weiss DJ, Cameron E, Bisanzio D, Mappin B, Dalrymple U, et al. The effect of malaria control on Plasmodium falciparum in Africa between 2000 and 2015. Nature. 2015;\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLwetoijera DW, Harris C, Kiware SS, Dongus S, Devine GJ, McCall PJ, et al. Increasing role of Anopheles funestus and Anopheles arabiensis in malaria transmission in the Kilombero Valley, Tanzania. Malar J. 2014;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaindoa EW, Matowo NS, Ngowo HS, Mkandawile G, Mmbando A, Finda M, et al. Interventions that effectively target Anopheles funestus mosquitoes could significantly improve control of persistent malaria transmission in south\u0026ndash;eastern Tanzania. PLoS One [Internet]. Public Library of Science; 2017;12:e0177807. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0177807\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0177807\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDjamouko-Djonkam L, Nkahe DL, Kopya E, Talipouo A, Ngadjeu CS, Doumbe-Belisse P, et al. Implication of Anopheles funestus in malaria transmission in the city of Yaound\u0026eacute;, Cameroon TT - Implication d\u0026rsquo;Anopheles funestus dans la transmission du paludisme dans la ville de Yaound\u0026eacute; au Cameroun. Parasite [Internet]. 2020/02/12. EDP Sciences; 2020;27:10. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pubmed.ncbi.nlm.nih.gov/32048986\u003c/span\u003e\u003cspan address=\"https://pubmed.ncbi.nlm.nih.gov/32048986\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOgola EO, Fillinger U, Ondiba IM, Villinger J, Masiga DK, Torto B, et al. Insights into malaria transmission among Anopheles funestus mosquitoes, Kenya. Parasit Vectors [Internet]. 2018;11:577. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13071-018-3171-3\u003c/span\u003e\u003cspan address=\"10.1186/s13071-018-3171-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMapua SA, Hape EE, Kihonda J, Bwanary H, Kifungo K, Kilalangongono M, et al. Persistently high proportions of plasmodium-infected Anopheles funestus mosquitoes in two villages in the Kilombero valley, South-Eastern Tanzania. Parasite Epidemiol Control [Internet]. 2022;18:e00264. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/science/article/pii/S2405673122000289\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/science/article/pii/S2405673122000289\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSwai JK, Mmbando AS, Ngowo HS, Odufuwa OG, Finda MF, Mponzi W, et al. Protecting migratory farmers in rural Tanzania using eave ribbons treated with the spatial mosquito repellent, transfluthrin. Malar J [Internet]. 2019;18:414. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-019-3048-8\u003c/span\u003e\u003cspan address=\"10.1186/s12936-019-3048-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatowo NS, Kulkarni MA, Messenger LA, Jumanne M, Martin J, Mallya E, et al. Differential impact of dual-active ingredient long-lasting insecticidal nets on primary malaria vectors: a secondary analysis of a 3-year, single-blind, cluster-randomised controlled trial in rural Tanzania. Lancet Planet Heal [Internet]. Elsevier; 2023;7:e370\u0026ndash;80. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S2542-5196(23)00048-7\u003c/span\u003e\u003cspan address=\"10.1016/S2542-5196(23)00048-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTakken W, Verhulst NO. Host Preferences of Blood-Feeding Mosquitoes. Annu Rev Entomol. 2013;58:433\u0026ndash;53.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLimwagu AJ, Kaindoa EW, Ngowo HS, Hape E, Finda M, Mkandawile G, et al. Using a miniaturized double-net trap (DN-Mini) to assess relationships between indoor\u0026ndash;outdoor biting preferences and physiological ages of two malaria vectors, Anopheles arabiensis and Anopheles funestus. Malar J [Internet]. 2019;18:282. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-019-2913-9\u003c/span\u003e\u003cspan address=\"10.1186/s12936-019-2913-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMidega JT, Mbogo CM, Mwnambi H, Wilson MD, Ojwang G, Mwangangi JM, et al. Estimating dispersal and survival of Anopheles gambiae and Anopheles funestus along the Kenyan coast by using mark-release-recapture methods. J Med Entomol [Internet]. 2007;44:923\u0026ndash;9. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.pubmedcentral.nih.gov/articlerender.\u003c/span\u003e\u003cspan address=\"http://www.pubmedcentral.nih.gov/articlerender.\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003efcgi?artid=2705338\u0026amp;tool=pmcentrez\u0026amp;rendertype=abstract\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoetzee M, Koekemoer LL. Molecular systematics and insecticide resistance in the major African malaria vector Anopheles funestus. Annu Rev Entomol. Annual Reviews; 2013;58:393\u0026ndash;412.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNambunga IH, Ngowo HS, Mapua SA, Hape EE, Msugupakulya BJ, Msaky DS, et al. Aquatic habitats of the malaria vector Anopheles funestus in rural south-eastern Tanzania. Malar J [Internet]. 2020;19:219. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-020-03295-5\u003c/span\u003e\u003cspan address=\"10.1186/s12936-020-03295-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDetinova TS. Age-grouping methods in Diptera of medical importance with special reference to some vectors of malaria. Monogr Ser World Health Organ. Geneva: World Health Organization; 1962;\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRao V. On Gonotrophic Discordance among certain Indian Anopheles. Indian J Malariol. 1947;1:43\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDerek Charlwood J, Nenhep S, Sovannaroth S, Morgan JC, Hemingway J, Chitnis N, et al. \u0026lsquo;Nature or nurture\u0026rsquo;: survival rate, oviposition interval, and possible gonotrophic discordance among South East Asian anophelines. Malar J [Internet]. 2016;15:356. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-016-1389-0\u003c/span\u003e\u003cspan address=\"10.1186/s12936-016-1389-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCharlwood JD, Tom\u0026aacute;s EVE, Andegiorgish AK, Mihreteab S, LeClair C. \u0026lsquo;We like it wet\u0026rsquo;: a comparison between dissection techniques for the assessment of parity in Anopheles arabiensis and determination of sac stage in mosquitoes alive or dead on collection. PeerJ. PeerJ Inc.; 2018;6:e5155.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOhm JR, Baldini F, Barreaux P, Lefevre T, Lynch PA, Suh E, et al. Rethinking the extrinsic incubation period of malaria parasites. Parasit Vectors [Internet]. 2018;11:178. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13071-018-2761-4\u003c/span\u003e\u003cspan address=\"10.1186/s13071-018-2761-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen CS, Mulla MS, March RB, Chaney JD. Cuticular hydrocarbon patterns in Culex quinquefasciatus as influenced by age, sex, and geography. Bull Soc Vector Ecol. 1990;15:129\u0026ndash;39.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCook PE, Hugo LE, Iturbe-Ormaetxe I, Williams CR, Chenoweth SF, Ritchie SA, et al. Predicting the age of mosquitoes using transcriptional profiles. Nat Protoc [Internet]. 2007;2:2796\u0026ndash;806. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nprot.2007.396\u003c/span\u003e\u003cspan address=\"10.1038/nprot.2007.396\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHugo LE, Cook PE, Johnson PH, Rapley LP, Kay BH, Ryan PA, et al. Field Validation of a Transcriptional Assay for the Prediction of Age of Uncaged Aedes aegypti Mosquitoes in Northern Australia. PLoS Negl Trop Dis [Internet]. Public Library of Science; 2010;4:e608. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pntd.0000608\u003c/span\u003e\u003cspan address=\"10.1371/journal.pntd.0000608\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang M-H, Marinotti O, James AA, Walker E, Githure J, Yan G. Genome-Wide Patterns of Gene Expression during Aging in the African Malaria Vector Anopheles gambiae. PLoS One [Internet]. Public Library of Science; 2010;5:e13359. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0013359\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0013359\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBurns DA, Ciurczak EW. Handbook of near-infrared analysis. CRC press; 2008.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayagaya VS, Michel K, Benedict MQ, Killeen GF, Wirtz RA, Ferguson HM, et al. Non-destructive determination of age and species of Anopheles gambiae sl using near-infrared spectroscopy. Am J Trop Med Hyg. 2009;81:622.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOng OTW, Kho EA, Esperan\u0026ccedil;a PM, Freebairn C, Dowell FE, Devine GJ, et al. Ability of near-infrared spectroscopy and chemometrics to predict the age of mosquitoes reared under different conditions. Parasit Vectors [Internet]. 2020;13:160. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13071-020-04031-3\u003c/span\u003e\u003cspan address=\"10.1186/s13071-020-04031-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSikulu-Lord MT, Maia MF, Milali MP, Henry M, Mkandawile G, Kho EA, et al. Rapid and Non-destructive Detection and Identification of Two Strains of Wolbachia in Aedes aegypti by Near-Infrared Spectroscopy. PLoS Negl Trop Dis. 2016;10.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLambert B, Sikulu-Lord MT, Mayagaya VS, Devine G, Dowell F, Churcher TS. Monitoring the Age of Mosquito Populations Using Near-Infrared Spectroscopy. Sci Rep. 2018;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSikulu-Lord MT, Devine GJ, Hugo LE, Dowell FE. First report on the application of near-infrared spectroscopy to predict the age of Aedes albopictus Skuse. Sci Rep. 2018;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSikulu M, Killeen GF, Hugo LE, Ryan PA, Dowell KM, Wirtz RA, et al. Near-infrared spectroscopy as a complementary age grading and species identification tool for African malaria vectors. Parasit Vectors. 2010;3:49.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSikulu MT, Majambere S, Khatib BO, Ali AS, Hugo LE, Dowell FE. Using a Near-Infrared Spectrometer to Estimate the Age of Anopheles Mosquitoes Exposed to Pyrethroids. PLoS One [Internet]. Public Library of Science; 2014;9:e90657. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0090657\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0090657\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGonzalez-Jimenez M, Babayan SA, Khazaeli P, Doyle M, Walton F, Reedy E, et al. Prediction of malaria mosquito species and population age structure using mid-infrared spectroscopy and supervised machine learning. Wellcome Open Res. 2019;4:76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSiria DJ, Sanou R, Mitton J, Mwanga EP, Niang A, Sare I, et al. Rapid age-grading and species identification of natural mosquitoes for malaria surveillance. Nat Commun [Internet]. 2022;13:1501. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-022-28980-8\u003c/span\u003e\u003cspan address=\"10.1038/s41467-022-28980-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMwanga EP, Siria DJ, Mitton J, Mshani IH, Gonz\u0026aacute;lez-Jim\u0026eacute;nez M, Selvaraj P, et al. Using transfer learning and dimensionality reduction techniques to improve generalisability of machine-learning predictions of mosquito ages from mid-infrared spectra. BMC Bioinformatics [Internet]. 2023;24:11. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12859-022-05128-5\u003c/span\u003e\u003cspan address=\"10.1186/s12859-022-05128-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMwanga EPP, Mapua SAA, Siria DJJ, Ngowo HSS, Nangacha F, Mgando J, et al. Using mid-infrared spectroscopy and supervised machine-learning to identify vertebrate blood meals in the malaria vector, Anopheles arabiensis. Malar J [Internet]. 2019;18:187. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-019-2822-y\u003c/span\u003e\u003cspan address=\"10.1186/s12936-019-2822-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMwanga EP, Minja EG, Mrimi E, Jim\u0026eacute;nez MG, Swai JK, Abbasi S, et al. Detection of malaria parasites in dried human blood spots using mid-infrared spectroscopy and logistic regression analysis. Malar J. 2019;\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoetzee M. Key to the females of Afrotropical Anopheles mosquitoes (Diptera: Culicidae). Malar J [Internet]. 2020;19:70. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-020-3144-9\u003c/span\u003e\u003cspan address=\"10.1186/s12936-020-3144-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoekemoer LL, Kamau L, Hunt RH, Coetzee M. A cocktail polymerase chain reaction assay to identify members of the Anopheles funestus (Diptera: Culicidae) group. Am J Trop Med Hyg [Internet]. 2002;66:804\u0026ndash;11. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.ncbi.nlm.nih.gov/entrez/query.\u003c/span\u003e\u003cspan address=\"http://www.ncbi.nlm.nih.gov/entrez/query.\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003efcgi?cmd=Retrieve\u0026amp;db=PubMed\u0026amp;dopt=Citation\u0026amp;list_uids=12224596\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePedregosa F, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res. 2011;\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbadi M, Barham P, Chen J, Chen Z, Davis A, Dean J, et al. TensorFlow: A system for large-scale machine learning. Proc 12th USENIX Symp Oper Syst Des Implementation, OSDI 2016. 2016.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChollet F. Keras: The Python Deep Learning library. KerasIo. 2015;\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMwanga EP, Siria DJ, Mitton J, Mshani IH, Jimenez MG, Selvaraj P, et al. Using transfer learning and dimensionality reduction techniques to improve generalisability of machine-learning predictions of mosquito ages from mid-infrared spectra. bioRxiv [Internet]. 2022;2022.07.26.501594. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://biorxiv.org/content/early/2022/\u003c/span\u003e\u003cspan address=\"http://biorxiv.org/content/early/2022/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e07/28/2022.07.26.501594.abstract\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNitish S, Geoffrey H, Alex K, Ilya S, Ruslan S. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. J Mach Learn Res. 2014;\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrechelt L. Early stopping - But when? Lect Notes Comput Sci (including Subser Lect Notes Artif Intell Lect Notes Bioinformatics). 2012;\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSuarez E, Nguyen HP, Ortiz IP, Lee KJ, Kim SB, Krzywinski J, et al. Matrix-assisted laser desorption/ionization-mass spectrometry of cuticular lipid profiles can differentiate sex, age, and mating status of Anopheles gambiae mosquitoes. Anal Chim Acta [Internet]. 2011;706:157\u0026ndash;63. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.sciencedirect.com/science/article/pii/S0003267011011639\u003c/span\u003e\u003cspan address=\"http://www.sciencedirect.com/science/article/pii/S0003267011011639\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDia I, Guelbeogo MW, Ayala D. Advances and Perspectives in the Study of the Malaria Mosquito Anopheles funestus. In: Manguin S, editor. Rijeka: IntechOpen; 2013. p. Ch. 7. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5772/55389\u003c/span\u003e\u003cspan address=\"10.5772/55389\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNtamatungiro AJ, Mayagaya VS, Rieben S, Moore SJ, Dowell FE, Maia MF. The influence of physiological status on age prediction of Anopheles arabiensis using near infra-red spectroscopy. Parasites and Vectors [Internet]. 2013;6:298. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://parasitesandvectors.biomedcentral.com/articles/\u003c/span\u003e\u003cspan address=\"http://parasitesandvectors.biomedcentral.com/articles/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/1756-3305-6-298\u003c/span\u003e\u003cspan address=\"10.1186/1756-3305-6-298\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"parasites-and-vectors","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"parv","sideBox":"Learn more about [Parasites \u0026 Vectors](http://parasitesandvectors.biomedcentral.com/)","snPcode":"13071","submissionUrl":"https://submission.nature.com/new-submission/13071/3","title":"Parasites \u0026 Vectors","twitterHandle":"@bugbittentweets","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Malaria, Anopheles funestus, deep learning, machine learning, Ifakara health institute, mid-infrared Spectroscopy","lastPublishedDoi":"10.21203/rs.3.rs-3834184/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3834184/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAccurately determining the age and survival probabilities of adult mosquitoes is crucial for understanding parasite transmission, evaluating the effectiveness of control interventions and assessing disease risk in communities. This study was aimed to demonstrating rapid identification of epidemiologically relevant age categories of \u003cem\u003eAnopheles funestus\u003c/em\u003e, a major Afro-tropical malaria vector, through the innovative combination of infrared spectroscopy and machine learning, instead of the cumbersome practice of dissecting mosquito ovaries to estimate age based on parity status.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAn. funestus\u003c/em\u003e larvae were collected in rural south-Eastern Tanzania and reared in the insectary. Emerging adult females were sorted by age (1–16 day-olds) and preserved using silica gel. PCR confirmation was conducted using DNA extracted from mosquito legs to verify the presence of \u003cem\u003eAn. funestus\u003c/em\u003e and eliminate undesired mosquitoes. Mid-infrared spectra were obtained by scanning the heads and thoraces of the mosquitoes using an ATR FT-IR spectrometer. The spectra (N = 2084) were divided into two epidemiologically relevant age groups: 1–9 days (young, non-infectious) and 10–16 days (old, potentially infectious). The dimensionality of the spectra was reduced using principal component analysis, then a set of machine learning and multi-layer perceptron (MLP) models were trained using the spectra to predict the mosquito age categories.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe best performing model, XGBoost, achieved an overall accuracy of 87%, with classification accuracies of 89% for young and 84% for old \u003cem\u003eAn. funestus\u003c/em\u003e. When the most important spectral features influencing the model performance were selected to train a new model, the overall accuracy increased slightly to 89%. The MLP model, utilising the significant spectral features, achieved higher classification accuracies of 95% and 94% for the young and old \u003cem\u003eAn. funestus\u003c/em\u003e, respectively. After dimensionality reduction, the MLP achieved 93% accuracy for both age categories.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study shows how machine learning can quickly classify epidemiologically relevant age groups of \u003cem\u003eAn. funestus\u003c/em\u003e based on their mid-infrared spectra. Having been previously applied to \u003cem\u003eAn. gambiae, An. arabiensis\u003c/em\u003e and \u003cem\u003eAn. coluzzii\u003c/em\u003e, this demonstration on \u003cem\u003eAn. funestus\u003c/em\u003e underscore the potential of this low-cost, reagent-free technique for widespread use on all the major Afro-tropical malaria vectors. Future research should demonstrate how such machine-derived age classifications in field collected mosquitoes correlate with malaria in human populations.\u003c/p\u003e","manuscriptTitle":"Rapid classification of epidemiologically relevant age categories of the malaria vector, Anopheles funestus","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-01-08 08:24:19","doi":"10.21203/rs.3.rs-3834184/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-02-18T15:10:52+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-02-02T00:49:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"dfde1cb2-0eb0-4276-9235-3f1286a79250","date":"2024-01-20T19:13:48+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"d48881e4-ede1-4b52-b599-2a677c1f9b28","date":"2024-01-17T09:12:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"3602e852-2b88-4260-934f-c0578ba10ffc","date":"2024-01-16T06:03:00+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-01-14T14:47:26+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-01-04T13:24:45+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-01-04T13:17:14+00:00","index":"","fulltext":""},{"type":"submitted","content":"Parasites \u0026 Vectors","date":"2024-01-04T09:36:07+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"parasites-and-vectors","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"parv","sideBox":"Learn more about [Parasites \u0026 Vectors](http://parasitesandvectors.biomedcentral.com/)","snPcode":"13071","submissionUrl":"https://submission.nature.com/new-submission/13071/3","title":"Parasites \u0026 Vectors","twitterHandle":"@bugbittentweets","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"90290581-0b09-4038-9ba2-734ee73268cc","owner":[],"postedDate":"January 8th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2024-03-25T15:05:24+00:00","versionOfRecord":{"articleIdentity":"rs-3834184","link":"https://doi.org/10.1186/s13071-024-06209-5","journal":{"identity":"parasites-and-vectors","isVorOnly":false,"title":"Parasites \u0026 Vectors"},"publishedOn":"2024-03-18 15:01:03","publishedOnDateReadable":"March 18th, 2024"},"versionCreatedAt":"2024-01-08 08:24:19","video":"","vorDoi":"10.1186/s13071-024-06209-5","vorDoiUrl":"https://doi.org/10.1186/s13071-024-06209-5","workflowStages":[]},"version":"v1","identity":"rs-3834184","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3834184","identity":"rs-3834184","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}