As part of “The Future of Forensic Analysis” content series presented by Spectroscopy, we sat down with Dr. Rajinder Singh of Department of Forensic Science, Punjabi University, Patiala, to talk about his recent work using attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FT-IR) to distinguish different animal species based on hair samples.
India is home to some of the most exotic animals, including the Royal Bengal tiger, Indian Leopard, and Snow Leopard. However, their natural habitats and livelihoods are under constant threat from poachers. Because these animals are used to create many valuable products on the black market, they are disproportionately targeted (1).
With the increase of wildlife crimes depleting the Bengal tiger, Indian Leopard, and Snow Leopard populations, the threat of species survival and loss of ecological balance becomes more plausible. As the Indian government and wildlife conservation organizations seek solutions to this growing problem, current efforts are underway to investigate wildlife crimes, and identifying animal species from hair has become an important component of pursuing prosecutions of those who commit wildlife crimes (1,2).
Rajinder Singh, Professor and former Head of the Department of Forensic Science, Punjabi University in Patiala, Punjab, India, is exploring how spectroscopy can help aid the forensic analysis of animal species and their hair samples. Recently, Singh and his team published a research paper titled, “Species Discrimination from Hair Using ATR FT-IR Spectroscopy: Application in Wildlife Forensics,” where they used attenuated total reflectance Fourier transform infrared (ATR FT-IR) spectroscopy associated with partial least squares discriminant analysis (PLS-DA) to help differentiate between animal species’ hair (1).
We sat down with Dr. Singh to discuss his research, and how spectroscopic techniques are being used in forensic analysis to help investigate and prosecute wildlife crimes.
Can you explain the significance of using ATR FT-IR spectroscopy for identifying hair samples in wildlife crimes?
ATR FT-IR spectroscopy has several advantages over other analytical techniques, including its non-destructive nature, ease of use, rapidity, and cost-effective analysis of samples with little to no sample preparation steps, thereby making it an ideal choice for examining hair and other biological evidence in wildlife crimes. ATR FT-IR spectroscopy is an environmentally friendly approach because no hazardous chemicals are utilized during sample analysis. In combination with chemometric tools, it offers a significant level of differentiation, provides objectivity, and eliminates biases from the results.
What are the key advantages of using a non-destructive technique like ATR FT-IR spectroscopy in forensic analysis?
Because trace amounts of evidence are frequently found at scene of crime, a non-destructive technique that require a small sample size will be preferred. Moreover, hair samples remain intact after ATR FT-IR spectroscopic analysis and can be used for any further examination including DNA barcoding.
Could you describe the chemometric models used in your study and how they contributed to species identification?
In this study, two chemometric models, principal component analysis (PCA) and PLS-DA were employed to identify and differentiate wild cat species. PCA is an unsupervised chemometric approach that reduces the dimensionality of correlated variables to a small number of significant variables, known as principal components (PCs). Combinations of these PCs are used to identify trends and patterns in the data set. As a result, samples with similar chemical compositions were gathered together, whereas samples with different compositions were separated apart. In this study, PCA model failed to separate hair samples of three wild cat species. Thus, PLS-DA, an advanced supervised chemometric tool, was employed. PLS-DA combines the characteristics of partial least square regression (PLSR) for reducing data dimensionality by selecting appropriate variables that maximize the covariance between the X and Y matrix with discriminant analysis to classify samples based on those derived variables. PLS-DA successfully separated hair samples of three wild cat species into three distinct classes.
What are PLS2-DA and PLS2DA-V (partial least squares discriminant analysis) models, and how do they differ in terms of algorithm functions, performance, and application?
To construct the PLS-DA model, two algorithms can be utilized such as PLS1-DA algorithms if the data set contains two classes and PLS2-DA algorithms if the data set contains more than two classes. The PLS2-DA algorithm was utilized to generate the PLS2-DA and PLS2DA-V models in this study because there were more than two classes. The selection of X-variables (spectral data points) was the primary distinction between the PLS2-DA and PLS2DA-V models. The PLS2-DA model used the full data set, consisting of 1666 spectral data points, for developing the model, whereas 1141 spectral data points with a VIP score ≥ 1 were used to develop the PLS2DA-V model. The performance of both models was compared using specificity, sensitivity, accuracy rate, recall, precision, and F1 score values, and it was discovered that the PLS2-DA model outperformed the PLS2DA-V model in identifying and discriminating three wild cat hair samples. This could be because of the combined contribution of information from other spectral data points with VIP scores < 1.
How does this study ensure the accuracy and precision of the PLS2-DA model during cross-validation and external validation?
The accuracy and precision of the PLS2-DA model during cross-validation and external validation were ensured by examining certain parameters such as R-squared, RMSE, and slope value. During cross-validation and external validation, high R-squared values and low RMSE values were found, indicating the prediction accuracy of the model with a low error rate. The slope values during cross-validation and external validation were near one, indicating that the data has been accurately modeled.
What role do VIP scores (variable importance in projection) play in the construction of chemometric models, and why was a threshold of ≥ 1 chosen in this study?
The VIP score assists in identifying the X-variables (wavenumbers) that make a substantial contribution to the development of the PLS-DA model. X-variables with VIP score ≥ 1 were chosen because the average value of the square of VIP scores is 1. The PLS-DA model constructed using selected X-variables with VIP score ≥ 1 can increase the performance of the model.
Can you elaborate on the process and significance of the blind test conducted with 10 unknown hair samples?
A blind test was conducted to determine and evaluate how well the model predicted the unknown samples. In our study, 10 unknown hair samples of three wild cats were provided to the analyst and their identity was not disclosed to him before the prediction was completed. These samples were analyzed using the same experimental conditions and parameters used for the training data set. The class of unknown samples was predicted using the developed PLS-DA model. The predicted Y scores of unknown samples were compared to the dummy Y-variables assigned to the known class, and an unknown sample was allocated to a known class if its Y-predicted score was equal or close to the Y-variable of that class.
How do the R-squared values for calibration and validation reflect the model’s performance in differentiating between the three wild cat species?
R-squared, also known as the coefficient of determination, is a statistical measure used in regression models to evaluate how much of the variance in the dependent variable can be explained by the independent variable. Its values range from 0 to 1, where 0 indicates that the model explains 0% of the correlation between the dependent and independent variables, and 1 indicates that the model explains 100% of the correlation. A high R-squared value (0.99 for calibration and 0.89 for validation) indicates the high goodness of fit of the model.
What challenges did you face during the implementation of this study, and how were they addressed?
Sample collection was a challenging task as these types of samples were not readily available. Therefore, all the samples for this study were obtained from the repository of wildlife forensic cell, the Wildlife Institute of India (WII), in Dehradun, India.
How might this method be applied in future wildlife forensic investigations, or other forensic investigations, and what potential improvements could be made?
ATR FT-IR spectroscopy, when combined with chemometrics, is gaining popularity among forensic scientists for the non-destructive examination of various sorts of wildlife or other forensic evidence. This technique employed an extremely small sample quantity with negligible to no sample preparation steps, which can facilitate the analysis of evidence found in trace amounts at the crime scene. One possible improvement that might be made to speed up and improve the accuracy of evidence analysis is the creation of a reference ATR FT-IR spectral database.
(1) Bhatia, D.; Sharma, C. P.; Sharma, S.; Singh, R. Species Discrimination from Hair Using ATR-FTIR Spectroscopy: Application in Wildlife Forensics. Science and Justice 2024, 64 (3), 314–321. DOI: 10.1016/j.scijus.2024.04.002
(2) Agnew, D. J.; Pearce, J.; Pramod, G.; et al. Estimating the Worldwide Extent of Illegal Fishing. PLoS One 2009; 4 (2), e4570. DOI: 10.1371/journal.pone.0004570
The Advantages and Landscape of Hyperspectral Imaging Spectroscopy
December 9th 2024HSI is widely applied in fields such as remote sensing, environmental analysis, medicine, pharmaceuticals, forensics, material science, agriculture, and food science, driving advancements in research, development, and quality control.
The Fundamental Role of Advanced Hyphenated Techniques in Lithium-Ion Battery Research
December 4th 2024Spectroscopy spoke with Uwe Karst, a full professor at the University of Münster in the Institute of Inorganic and Analytical Chemistry, to discuss his research on hyphenated analytical techniques in battery research.
AI, Deep Learning, and Machine Learning in the Dynamic World of Spectroscopy
December 2nd 2024Over the past two years Spectroscopy Magazine has increased our coverage of artificial intelligence (AI), deep learning (DL), and machine learning (ML) and the mathematical approaches relevant to the AI topic. In this article we summarize AI coverage and provide the reference links for a series of selected articles specifically examining these subjects. The resources highlighted in this overview article include those from the Analytically Speaking podcasts, the Chemometrics in Spectroscopy column, and various feature articles and news stories published in Spectroscopy. Here, we provide active links to each of the full articles or podcasts resident on the Spectroscopy website.
Mass Spectrometry for Forensic Analysis: An Interview with Glen Jackson
November 27th 2024As part of “The Future of Forensic Analysis” content series, Spectroscopy sat down with Glen P. Jackson of West Virginia University to talk about the historical development of mass spectrometry in forensic analysis.