Researchers at Jiangsu University of Science and Technology showed that FT-IR spectroscopy combined with optimized chemometric modeling can rapidly and accurately detect and stage Bombyx mori nucleopolyhedrovirus infection in silkworms.
According to a team of researchers based out of Jiangsu University of Science and Technology, combining chemometric modeling with Fourier transform infrared (FT-IR) spectroscopy can greatly improve accuracy when detecting Bombyx mori nucleopolyhedrovirus (BmNPV) infection in silkworms. Published in the journal Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, the study demonstrates how this method could conduct rapid, reagent-free screening using small hemolymph volumes, addressing a major diagnostic bottleneck in commercial sericulture.1
What is BmNPV?
BmNPV is a pathogenic virus that primarily impacts silkworms.2 According to estimates that the researchers provided, BmNPV accounts for roughly 70% of annual disease losses in silkworm farming worldwide, making early detection critical for containment and productivity.1 Another complication is that the immune mechanisms of this disease are not certain.2 Currently, detection methods for BmNPV mostly use PCR-based or immunological conventional assays. The problem is that these methods require laboratory infrastructure, trained personnel, and time-consuming sample preparation.1
What did the researchers do in their study?
In their study, the researchers demonstrated how their new FT-IR–chemometrics workflow overcomes these obstacles. By acquiring the mid-infrared (MIR) spectra directly from silkworm hemolymph, the method allowed the team to conduct rapid analysis compatible with field deployment.1 For sericulture operations and diagnostic laboratories, the work suggests a pathway toward on-site infection screening and staging, potentially reducing outbreak spread and economic losses.1
As part of the experimental procedure, the team analyzed 840 hemolymph samples collected across five days post-infection.1 When they conducted the initial unsupervised principal component analysis (PCA), the researchers observed substantial overlap among infection stages. This observation indicated that raw spectral differences alone could not reliably separate healthy from infected states.1 In an attempt to remedy this, the researchers tried to improve classification by using linear discriminant analysis (LDA). However, this method still produced errors.1
Therefore, the researchers trialed nonlinear and latent-variable models after spectral preprocessing. A k-nearest neighbors (kNN) classifier following baseline correction and mean centering reached 99.29% overall accuracy, with sensitivity and specificity both exceeding 99%.1 The highest performance came from partial least squares-discriminant analysis (PLS-DA) using first-derivative and mean-centered spectra, which achieved perfect classification in both the calibration and prediction sets.1
Beyond binary detection, PLS-DA latent variable plots mapped the temporal progression of BmNPV infection. Score trajectories along the first two latent variables (LV1 = 19.71%; LV2 = 23.44%) corresponded to metabolic shifts from early energy mobilization to late-stage systemic collapse.1 This capability to stage infection non-destructively could support targeted interventions, such as isolating early-stage carriers before colony-wide spread.1
A unified spectral pretreatment workflow combining first-derivative transformation with standard normal variate (SNV) correction minimized non-infection-related spectral variability. According to the authors, this step was essential to remove baseline drift and scattering effects intrinsic to biological samples, allowing chemometric models to capture infection-specific biochemical changes.1 The approach was implemented with attenuated total reflectance FT-MIR (ATR-FT-MIR) acquisition, which is a configuration commonly available in industrial and academic spectroscopy laboratories.1
For analytical chemists and instrument developers, the study highlights how preprocessing choices can outweigh algorithm selection in biological classification tasks. The comparison across PCA, LDA, kNN, and PLS-DA models underscores the importance of integrating optimized signal correction with appropriate multivariate modeling to reach deployable accuracy levels.1
What are the implications for sericulture and applied spectroscopy?
BmNPV can wreak havoc on silk production. That is why it is paramount that those in the industry have access to routine screening methods that are accurate. A portable or benchtop FT-IR diagnostic that requires minimal consumables could enable routine screening at rearing facilities rather than centralized laboratories.1 The authors note that early detection and staging support timely quarantine and treatment decisions, improving sustainability and yield in silk production systems.1
Because FT-IR instruments are already used for quality control in many industries, adapting them for sericulture diagnostics may be technically feasible without major infrastructure changes.1 For sericulture stakeholders and applied spectroscopy professionals, this represents a shift toward practical, on-site viral screening using existing spectroscopic hardware, potentially lowering diagnostic costs and improving disease management in silk production.
