Current Status and Future Perspectives of Human Saliva Identification in Forensic Science
2026 年 14 巻 1 号 p. 136-152
詳細
2026 年 14 巻 1 号 p. 136-152
DNA typing is a powerful tool in forensic science; however, its evidentiary value depends on clarifying the origin and context of DNA deposition. Within the hierarchical structure of criminal proof, DNA typing typically addresses the sub-source level and requires the identification of the associated body fluid to strengthen its relevance. Among the body fluids found at crime scenes, including blood, semen, and saliva, saliva is particularly significant because of its high retention rate and broad applicability beyond violent or sexual crimes. It can be recovered from bite marks, cigarettes, food remnants, envelopes, and even vehicle components and may appear through indirect transfer, such as coughing or sneezing. Saliva-derived DNA often plays a major role in forensic identification, offering strong potential for individualization. This review explores the properties of human saliva, current detection methods, and future perspectives, emphasizing its critical role in enhancing the interpretative power of DNA evidence.
In recent years, DNA typing has become a widely used tool to establish the factual basis of criminal investigations. Although DNA recovered from crime scene evidence has high evidentiary value, it is often insufficient to establish the facts of a crime conclusively. This limitation arises from the need to clarify how and under what conditions DNA is deposited in crime scene evidence [1].
In general, the proof of a crime follows a hierarchical structure (Fig. 1): Establishing an offense requires demonstrating an act, and demonstrating an act requires identifying its source. Each source may consist of sub-sources, and evidence at each level must be supported by evidence from lower levels. For example, in murder cases, the prosecution must prove that the suspect actively stabbed the victim using a knife. To prove the act of stabbing, the blood on the knife must be proven to originate from the victim, and this source must be linked to a sub-source whose DNA profile matches that of the victim. Within this hierarchy, even when a complete profile is obtained, DNA typing remains at the sub-source level. It can only become a source when the associated body fluid or tissue is identified. In the aforementioned murder case, obtaining the victim's DNA type from the red liquid on the knife alone is insufficient; only when the liquid is conclusively proven to be human blood can it serve as a source of evidence. Therefore, it is crucial to determine the specific body fluids present in crime scene evidence.
Body fluids commonly encountered at crime scenes include blood, semen, and saliva. Blood is frequently left behind in violent crimes such as murder, while semen is critical evidence in sexual assault cases. However, blood and semen are less likely to be retained in crimes such as theft, which constitute most crimes in Japan [2]. In contrast, the saliva exhibits a much higher retention rate. It can be recovered from bite marks [3], cigarettes [4], leftover food [5], envelopes and postage stamps [6, 7], and airbags of cars involved in accidents [8], making it important evidence in cases beyond violent or sexual crimes. Additionally, saliva can spread through coughing or sneezing, allowing for its presence even in locations where direct oral contact is unlikely. For example, saliva can adhere to automobile steering wheels, gear sticks, and consoles [9].
Moreover, saliva-derived DNA is often a major contributor to DNA typing [10] and has high evidentiary value owing to its suitability for individual identification. Consequently, the evidentiary value of saliva is comparable, if not superior, to that of blood or semen because of its extensive potential for retention at crime scenes.
By integrating the fundamental biological properties of human saliva with current analytical approaches, this review aims to clarify both the present capabilities and future perspectives of saliva identification in forensic science.
Human saliva is a complex fluid secreted by several salivary glands, including the parotid and submandibular glands. It plays vital roles in the oral cavity, including lubrication, oral protection, buffering, maintenance of tooth integrity, antibacterial activity, taste perception, and digestion [11, 12]. Blood production is approximately 20 mL/day [13], and a single semen ejaculation averages approximately 3.8 mL [14]. In contrast, saliva production is substantially higher, ranging from 0.5 to 1.5 liters per day [15], which is considerably higher than that of blood and semen. Despite its large volume, nearly all saliva is swallowed and does not exit the body because the presence of approximately 1.1 mL of saliva in the oral cavity triggers the swallowing reflex [15].
Additionally, salivary secretion is dynamic and can change readily. Under unstimulated conditions, secretion is 0.3–0.4 mL/min, decreasing to approximately 0.1 mL/min during sleep. However, stimulation by chewing, taste, or smell can markedly increase secretion to approximately 7 mL/min [15, 16]. Saliva production also varies with age [17], region of residence [18], certain diseases (e.g., obesity and periodontal disease) [19, 20], and lifestyle habits such as diet and smoking [21, 22]. Mental state also influences saliva production; it decreases during relaxation, such as when listening to music, but increases when watching anxiety-provoking videos [15].
Consequently, the amount of saliva left at a crime scene can vary widely, depending on the circumstances. For example, 10 mL of blood left at a crime scene suggests bodily injury, whereas 10 mL of semen may indicate multiple ejaculations. By contrast, 10 mL of saliva alone is unlikely to provide conclusive evidence. Moreover, estimating the exact amount of saliva from the DNA content is difficult because of diurnal and individual variations in salivary DNA concentrations [23].
Saliva is a dilute, colorless body fluid composed of over 99% water and contains electrolytes (sodium, potassium, calcium, and magnesium), carbonic acid, phosphoric acid, proteins, peptides, enzymes, mucins, immunoglobulins, urea, ammonia, and various other components [24]. Unlike blood and semen, saliva contains food debris and microorganisms [25]. To confirm the presence of saliva, it is necessary to analyze its characteristics that are not found in other organisms or body fluids. Therefore, in forensic science, the presence of saliva is typically determined by detecting components unique to human saliva and evaluating their ratios. Specifically, analytical methods focus on (1) proteins and peptides, (2) DNA and RNA, and (3) spectroscopic techniques, such as Raman spectroscopy and nuclear magnetic resonance. The following sections describe these methods in detail.
The protein- and peptide-based methods described in this section differ substantially in their analytical principles, specificity, and practical applicability in forensic investigation. To facilitate an overview and comparison of these approaches, their principles, advantages, limitations, and typical use scenarios are summarized in Table 1.
3.1 α-Amylase-based strategies (classical markers and activity assays)Saliva contains various proteins and peptides, with α-amylase being the most abundant [26]. This enzyme is present in human saliva and that of certain omnivorous animals but is absent in carnivores (e.g., dogs and cats) and herbivores (e.g., horses and cows), which are commonly encountered as pets or livestock [27]. Moreover, human saliva exhibits significantly higher α-amylase activity than other bodily fluids [28]. In forensic science, this property has been used to identify saliva by measuring α-amylase activity.
Historically, the iodine-starch complex reaction was used, where the purple color fades owing to α-amylase activity. Currently, the blue starch method is widely employed, and the Phadebas Amylase Test is a representative commercial product [29]. This method detects saliva by exploiting the reaction in which blue dye bound to starch is solubilized in the presence of α-amylase activity, and is also applied to visualize saliva spots on clothing and fabric. The original blue starch method involves a solution-based reaction, whereas the blue starch agarose plate method allows the reaction to occur directly within a gel matrix [30, 31]. In this approach, α-amylase activity is visualized by placing samples directly on a gel plate containing powdered blue starch. A major advantage of this method is its ability to simultaneously process multiple samples. However, for colored samples where blue color detection is difficult, commercially available kits such as SALIgAE® provide an alternative. This kit changes color to yellow in the presence of α-amylase, which facilitates the identification of saliva [32]. Recent research has explored small-molecule synthetic substrates instead of starch for α-amylase detection, including N3-G5-β-CNP (2-chloro-4-nitrophenyl 65-azido-65-deoxy-β-maltopentaoside) and Gal-G2-CNP (2-chloro-4-nitrophenyl-4-O-β-d- galactopyranosylmaltoside) [33, 34]. Additionally, the development of starch-iodine-coated nanoparticle sensors with a chitosan-tripolyphosphate core has been reported [35]. These nanoparticle sensors are blue under normal conditions but turn red when amylase disrupts the starch-iodine structure.
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/14_136_2.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
While saliva identification using α-amylase activity is a valid approach, it has notable limitations. α-Amylase activity is not exclusive to human saliva; it is also found in microorganisms, insects, and plants [27, 36, 37, 38]. Furthermore, the human body produces two types of α-amylases: salivary α-amylase (AMY1) and pancreatic α-amylase (AMY2) [39]. AMY1 is present in saliva, breast milk, and sweat, whereas AMY2 is present in seminal and vaginal fluids [40]. The α-amylase activity of semen is generally much lower than that of saliva, typically less than one-thousandth of salivary activity. However, some individuals may exhibit α-amylase activity in semen comparable to saliva [41]. Furthermore, certain vaginal microorganisms can produce α-amylase [42], meaning that vaginal fluid may contain high microbial α-amylase activity. Consequently, relying solely on α-amylase activity measurements for forensic saliva identification may be problematic. As summarized in Table 1, α-amylase activity assays remain primarily screening tools and should be complemented by more specific confirmatory methods.
Saliva, a transparent bodily fluid, is difficult to detect visually and is often challenging to collect at a crime scene. Therefore, in forensic practice, it is necessary to screen samples from locations where saliva is suspected and to test only those with a high likelihood of being saliva. For such screening applications, high discrimination is not essential; rather, simple and cost-effective processing of a large number of samples is critical. One example is the amylase activity assay using blue starch, which detects all α-amylase activity. A positive result does not definitively confirm the presence of human saliva. However, this method is extremely inexpensive and allows measurement by simply pressing a blue-starch-coated sheet (e.g., Phadebas® Forensic CR, Phadebas AB, Sweden) against the sample. This simplicity enables large-area searches for saliva on crime evidence, such as bedsheets, carpets, and clothing, a feature unmatched by other saliva identification methods. Consequently, α-amylase-based screening remains widely used despite its low specificity. If a method could be developed to improve specificity while maintaining simplicity and affordability, it could become a highly practical tool for forensic investigations. From this perspective, the method based on α-amylase activity also offers a straightforward explanation that is easy to communicate to non-specialists.
3.2 AMY1-specific identification (antibody-based and mass spectrometry approaches)To overcome the limitations in activity measurements, methods have been developed to identify human salivary α-amylases. These methods can be divided into three main categories: (1) enzymatic identification, (2) antibody-based identification, and (3) mass spectrometry. Enzymatic methods exploit differences in the reactivity of AMY1 and AMY2 toward specific inhibitors, such as those derived from kidney beans or wheat [43], and differences in the reaction kinetics between the two enzymes [44]. While these methods offer greater specificity than simple α-amylase activity measurements, their forensic applicability is limited owing to the potential presence of other α-amylases in samples and the complexity of testing. Consequently, direct AMY1 detection methods have been developed.
One such approach is antibody-based detection of AMY1. A classic example is the immunodiffusion assay [38], which relies on precipitate formation in a gel when AMY1 from a sample reacts with anti-AMY1 antiserum. This method involves creating two wells in a thin agarose gel on a glass plate; the sample is placed in one well and the anti-AMY1 antiserum in the other. In the presence of AMY1, precipitation lines are formed by diffusion. This technique offers a significant advantage over simple α-amylase activity measurement by enabling differentiation between human and rodent saliva, even when their α-amylase activities are similar. Enzyme-linked immunosorbent assay (ELISA) provides high sensitivity for AMY1 detection [45]. However, antisera-based methods have a potential limitation in that antibody specificity may vary substantially. For example, in the evaluation of antibody quality in epigenetics, one-quarter of 246 histone modification antibodies bound to two or more antigens, and another four bound specifically to an unintended antigen [46]. To overcome this issue, the RSID™-Saliva immunochromatography kit (Independent Forensic, Lombard, IL, USA) has gained widespread adoption [29]. This kit uses rigorously quality-controlled, lot-independent monoclonal antibodies, enabling specific AMY1 detection in saliva samples as small as 1 µL. Furthermore, this assay does not exhibit cross-reactivity with blood, semen, urine, or vaginal fluid [47], making it widely used in forensic laboratories. Nevertheless, antibody-based methods cannot detect all AMY1 isoforms because human saliva contains more than 65 AMY1 variants that differ in glycosylation and splicing [48].
Antibody-independent methods have been developed to overcome these limitations. The most direct approach is the matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF/MS) analysis of AMY1 purified with glycogen, which detected 55 charged ions [49]. However, as this method targets full-length AMY1, multiple MS standards are required to identify other AMY1 isoforms. To overcome this limitation, a proteomics method combining trypsin digestion and tandem MS (MS/MS) analyses was developed. This method identified AMY1 peptides using liquid chromatography (LC)/MS analysis and matched the obtained mass spectra with a database (Mascot analysis), enabling the detection of as little as 0.1 µL of saliva [50]. However, this method does not fully comply with the Guidelines for Forensic Science Appraisal (SWGTOX) [51] because peptide sequences are determined probabilistically. To meet these guidelines, a new method employs stable isotope-labeled AMY1 peptide standards for LC/MS analysis [52]. Trypsin-digested saliva samples are spiked with labeled standards, followed by simultaneous LC/MS analysis of both labeled and endogenous peptides. Compared to Mascot-based analysis, this method offers enhanced sensitivity and broader detection of AMY1 fragments, successfully identifying AMY1 from as little as 0.0244 µL of saliva.
3.3 Other protein/peptide markers and practical considerationsIn addition to AMY1, other salivary proteins and peptides can also serve as markers. One such marker is statherin, a 43-amino acid peptide that prevents calcium phosphate precipitation in the oral cavity [53] and is specifically expressed in salivary glands [54]. Statherin has been detected using ELISA [45] and LC/MS [52], making it a valuable indicator for saliva identification. Its relatively small molecular weight (~5.4 kDa) allows for intact mass analysis without batch operations, such as trypsin digestion [55], enabling streamlined and cost-effective saliva identification. Submaxillary gland androgen-regulated protein 3B (SMR3B) [54], another salivary gland-specific peptide, has also been used for saliva identification via LC/MS analysis [52]. Additionally, proline-rich proteins are saliva-specific, and an ELISA-based method for saliva identification has been developed [56]. Comprehensive MALDI-TOF/MS spectrometry profiling of small molecules combined with principal component analysis of the MS spectra has also been employed to discriminate saliva from other body fluids [57]. Although MS-based methods offer high sensitivity and specificity, their complexity and longer processing time compared with immunochromatography kits have limited their widespread use.
Nucleic acid-based approaches for saliva identification encompass a broad spectrum of strategies, including RNA expression profiling, epigenetic analysis, and microbiome-based identification. Although these methods offer high biological specificity, they differ markedly in stability requirements, DNA consumption, and practical feasibility. A comparative summary of their principles, advantages, limitations, and forensic applicability is presented in Table 2.
4.1 mRNA-based saliva identificationHuman saliva contains saliva-specific peptides, and the messenger RNAs (mRNAs) encoding these peptides are also present in saliva. Consequently, methods have been developed to detect these mRNAs in saliva. For example, mRNA encoding histatin, a 32-amino acid peptide with antimicrobial activity in the oral cavity, is present in saliva. Similarly, mRNA for statherin, a protein discussed previously, is also present in saliva. By detecting these mRNAs using reverse transcription polymerase chain reaction (RT-PCR), saliva can be distinguished from other body fluids (blood, semen, vaginal fluid, menstrual blood, and nasal secretions) [58, 59, 60, 61]. Reverse transcriptase loop-mediated isothermal amplification (RT-LAMP) is a convenient method for statherin mRNA detection [62, 63]. RT-LAMP enables the visual confirmation of mRNA amplification without requiring specialized equipment, making it potentially useful for field-based saliva identification. A more convenient method combines reverse transcription recombinase polymerase amplification (RT-RPA) with lateral flow dipstick assays [64]. RT-RPA is a low-temperature isothermal method that simultaneously performs reverse transcription and DNA amplification, requiring only simple handling, such as holding the tube by hand [65]. When combined with a lateral flow dipstick, mRNA-based testing can be performed in an environment without the use of electronic devices. Commercially available kits are ideal for routine identification of body fluids using mRNA detection. Currently, a kit for histatin mRNA detection is available for identifying fluid species [66].
While saliva identification is important for establishing the context of a crime, it is equally important to determine the individual to whom the saliva belongs. Therefore, saliva identification methods should not interfere with DNA typing, which remains the cornerstone of forensic analysis. Recent advancements include protocols for the simultaneous extraction of DNA and RNA, enabling body fluid identification through RNA analysis, while retaining DNA for typing. By co-extracting DNA and other components, and using only non-DNA components for saliva identification, competition with DNA typing can be minimized [67].
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/14_136_3.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
mRNAs are not the only endogenous RNAs present in the saliva; several other RNA types hold promise for saliva identification. One example is microRNAs (miRNAs), a class of small non-coding RNAs (20–25 bases) used as markers for body fluid identification [68]. Although saliva-specific miRNA markers remain to be discovered, multiple miRNA markers such as miR-200c-3p, miR-203a, and miR-205-5p can be detected using TaqMan-based quantitative reverse transcription PCR (RT-qPCR). The combination of the detection results of these markers may serve as a criterion for saliva identification [69]. Furthermore, several methods have been developed to classify body fluids by integrating multiple miRNAs detected via RT-qPCR and using algorithms based on continuous models [70, 71]. However, the accuracy of saliva discrimination using miRNAs can be very low (approximately 7%) due to variations in miRNA expression associated with recent food intake. Piwi-interacting RNAs (piRNAs), another class of non-coding RNAs similar to miRNAs, hold promise as body fluid markers. A saliva identification method that combines multiple piRNAs and applies algorithmic analysis has been developed [72]. Since body fluid-specific piRNA markers, such as piR-55521 in semen and piRNAs with differential expression across body fluids have been identified [73], the discovery of saliva-specific piRNA markers is highly anticipated. These markers would simplify the testing procedure. Similarly, the development of a saliva identification method using saliva-specific circular RNAs (circRNAs) has been anticipated. CircRNAs represent another class of non-coding RNAs present in saliva that exhibit considerable diversity in different body fluids [74]. In addition to circRNAs, other non-coding RNAs, including antisense RNAs (asRNA), small nuclear RNAs (snRNAs), small nucleolar RNAs (snoRNAs), competing endogenous RNAs (ceRNAs), and long intergenic ncRNAs (lincRNAs), are potential novel markers for saliva identification [75].
However, RNA-based saliva identification methods face challenges due to their inherent susceptibility to degradation. For example, β-actin mRNA is undetectable in dried bloodstains stored at room temperature for only 14 days when analyzed using RT-PCR [76]. Similarly, statherin or histatin 3 mRNA cannot be detected in saliva samples soaked in cotton cloth outdoors for 30 days, even without rain exposure [77]. Moreover, RNA in saliva is more prone to degradation than that in blood or semen [78]. As forensic samples may need to be preserved for extended periods (potentially decades) before analysis, rapidly degrading mRNAs are considered unsuitable for testing. As outlined in Table 2, RNA-based approaches are intrinsically constrained by degradation, highlighting the need for careful consideration of sample age and storage conditions. In contrast, circRNAs are generally more resistant to degradation than mRNAs [79], suggesting that future methods targeting stable RNAs such as circRNAs could overcome storage-related limitations.
4.3 Human DNA methylation-based identificationIn contrast to RNA-based methods summarized in Table 2, DNA methylation analysis benefits from the higher stability of DNA but requires careful management of DNA consumption. Although the sequence of somatic DNA is identical across all cells, methylation patterns vary among tissues [80]. This variation has been used to develop saliva identification methods. One such method employs the methylation-sensitive restriction enzyme Hha I [81, 82]. This enzyme recognizes the GCGC site and cleaves DNA unless the cytosine is methylated. Consequently, if a PCR primer site includes an Hha I recognition sequence and the cytosine is methylated, PCR amplification will not occur. By analyzing methylation patterns at loci containing Hha I sites, researchers can differentiate saliva from blood, semen, skin, and menstrual blood. A more specific method combines the methylation-sensitive enzyme Sma I with the methylation-dependent enzyme Gla I [83]. Although these methods can identify body fluids using as little as 1 ng of DNA, enzyme activity may compromise their accuracy.
Bisulfite-based methods have been developed to overcome these limitations. Bisulfite treatment converts unmethylated cytosines to uracil while leaving methylated cytosines unchanged. Subsequent sequencing techniques, such as pyrosequencing, can be used to identify the methylated regions. The breast carcinoma amplified sequence 4 (BCAS4) region in saliva is highly methylated compared with other body fluids, and analysis via bisulfite treatment and pyrosequencing enables saliva identification [84, 85]. Additional saliva-specific methylation regions, including hippocalcin-like protein 1 (HPCAL1), spindle assembly abnormal protein 6 homolog (SA-6), and stromal antigen 1 (SA1), have also been identified [86, 87, 88].
Another approach combines bisulfite treatment with single-nucleotide extension, allowing compatibility with the capillary electrophoresis systems commonly used in forensic laboratories [89]. Furthermore, comprehensive DNA methylation analysis using bead arrays has revealed that saliva exhibits higher methylation rates in the solute carrier family 12 member 8 (SLC12A8, cg26107890) and SOX2 overlapping transcript (SOX2OT, cg20691722) regions than blood, semen, and vaginal fluid [90], suggesting the potential for bead array-based saliva identification.
However, compared with conventional DNA typing methods, such as GlobalFiler, which require only 1 ng of template DNA [91], DNA methylation-based methods typically demand larger DNA quantities. As crime scene evidence may yield limited DNA, a balance between saliva identification and DNA typing must be considered for practical applications.
4.4 Oral microbiome-based identificationThe human body harbors a vast number of microorganisms, nearly as many as somatic cells [92], and the oral cavity is home to at least 700 distinct species [93]. Because the oral microbiome is stable [94, 95], saliva identification can be achieved by detecting the DNA of saliva-specific microorganisms. For instance, several bacterial species, including Streptococcus salivalius, S. oralis, S. sanguinis, Neisseria subflava, N. flavescens, Prevotella melaninogenica, and Fusobacterium nucleatum, are found almost exclusively in saliva [96, 97, 98]. Additionally, the DNA of S. mutans and Scardovia wiggsiae may serve as valuable markers for identifying saliva, particularly in individuals with dental caries [99]. The presence of these microorganisms can be confirmed by detecting the 16S rRNA using real-time PCR. Therefore, saliva identification methods targeting microorganisms can be performed in standard laboratories.
For a simpler and more affordable approach, a lateral flow dipstick detection method was developed to eliminate the need for a real-time PCR system [100]. Furthermore, a saliva identification method for S. salivarius that directly detects bacterial cells was developed. This method uses nanoparticles coated with the antibacterial peptide GH12. When these nanoparticles encounter S. salivarius, they aggregate, and subsequent fluorescence detection of this aggregate confirms this [101]. S. salivarius is more abundant in licking-derived (contact) saliva than in drooling-derived (non-contact) saliva. This observation has led to the development of a method for estimating "licking" activity by quantifying the amount of S. salivarius DNA [102].
One challenge with DNA-based saliva identification is that oral microorganisms can also be detected in feces as saliva passes through the gastrointestinal tract. To address this, a system has been developed that simultaneously detects saliva- and fecal-specific microorganisms, such as Bacteroides thetaiotaomicron and B. uniformis, to improve specificity [103]. Another challenge is distinguishing saliva from vaginal fluid and skin samples, which is particularly relevant in sexual assault investigations, where the perpetrator's saliva may be present on the victim's skin or vagina. To address this issue, a method based on deep learning of 16S rRNA data from the Human Microbiome Project was developed. This method analyzes the microbiome data of the skin, oral cavity, and vagina to determine the origin of a sample using an artificial neural network [104]. The first two principal components of this method can explain 92% of the data variation and clearly distinguish the samples from the skin, oral cavity, or vagina. However, in the mock casework, an oral sample stored for seven years was misclassified as of skin origin. Further validation is required, particularly for populations and samples other than those used in the original study.
For instance, when applied to a Japanese population, validation of the oral microflora is essential. Saliva samples from populations in Japan, the United States, Korea, Italy, and India were analyzed to calculate the genus-level occupancy of each microorganism, and those with an occupancy of 10% or higher were ranked [105]. The results showed that Streptococcus spp. ranked first or second in populations from the United States, Korea, Italy, and India, whereas Streptococcus spp. did not exceed 10% of the Japanese population. Moreover, significant differences in oral microflora were observed, even between geographically proximate areas in Japan. For instance, Bifidobacterium spp. and Blautia spp. are dominant in Hisayama Town, while Collinsella spp. and Granulicatella spp. are dominant in the adjacent Fukuoka City. Therefore, a large-scale validation is necessary to determine the effectiveness of microflora-based salivary identification in local populations in Japan.
Methods targeting microorganisms in saliva are intuitive for non-specialists. However, population structure and storage-related effects necessitate local validation and careful interpretation in casework.
Protein/peptide- and DNA/RNA-based methods for saliva identification offer high sensitivity and accuracy; however, they are inherently destructive because the sample is consumed during analysis. To address this limitation, researchers have developed nondestructive approaches that preserve samples for subsequent testing. Raman spectroscopy is one such method. Salivary amylase (AMY1), which contains the amino acid phenylalanine, exhibits a distinct Raman peak at 1000 cm-1, attributable to the aromatic ring breathing of the phenylalanine side chain. Additionally, saliva exhibits a unique spectral peak at 2065 cm-1, which is absent in blood, semen, vaginal fluid, and sweat [106]. These features enable discrimination of saliva from other body fluids using Raman spectra.
Body fluids at crime scenes often adhere to clothing, and direct Raman measurements of such samples result in mixed spectra of both fabric and body fluids. Recently, methods have been developed to separate Raman signals originating from clothing from those originating from body fluids using blood samples [107]. Applying this technique to saliva could enable the direct and nondestructive identification of saliva traces at crime scenes. Furthermore, the Raman spectrum of saliva remains stable for a short period (approximately seven days) but varies with the age of the individual. This property has been leveraged to estimate the age of saliva donors using algorithms based on artificial neural networks [108]. The current accuracy is approximately 80% for the three age categories (young: 20–30 years, mid: 31–55 years, and older: 56+ years). Further refinement of this approach could provide valuable age-related information for criminal investigations through nondestructive testing.
Another nondestructive method involves proton nuclear magnetic resonance (1H NMR) spectroscopy. 1H NMR analysis of saliva revealed a spectrum of metabolites, including lactic acid, propionic acid, acetic acid, and formic acid. When combined with specific algorithms, these metabolic signatures can be used to distinguish saliva from semen, serum, and urine [109]. However, this method currently requires 1 mL of centrifuged saliva supernatant, raising concerns regarding the detection sensitivity.
Fluorescence spectroscopy offers an additional approach for the nondestructive identification of saliva. Conventional body fluid detection at crime scenes typically relies on single-wavelength light sources [40], which limit the information obtained. A more advanced multispectral fluorescence imaging system was developed, using excitation light from 200 to 600 nm and analyzing the fluorescence emission from 220 to 700 nm [110]. This system applies principal component analysis to spectral data for saliva identification and demonstrates promising performance on mock samples.
Finally, Fourier-transform infrared (FT-IR) spectroscopy combined with partial least squares discriminant analysis (PLS-DA) was explored as a nondestructive method. This method analyzes FT-IR spectra using PLS-DA to differentiate saliva from other bodily fluids [111].
Nondestructive analyses (Raman, 1H NMR, fluorescence, FT-IR) preserve DNA for downstream typing and thus align well with the priority of compatibility with DNA typing. Practical constraints currently include sample size and the need for robust calibration/algorithms for varied substrates.
5.2 Machine learning and data integration (applications across modalities)Machine learning has increasingly been applied to the analysis of saliva-related forensic data to improve the discrimination and interpretation of complex biological signals. One representative application is age estimation, in which spectral features of saliva were analyzed using artificial neural networks to predict the age categories of saliva donors [108].
Another major challenge in saliva identification is the discrimination of saliva from biologically and anatomically related body fluids, particularly vaginal fluid and skin-derived traces. To address this issue, a deep learning-based approach using 16S rRNA sequencing data from the Human Microbiome Project was developed. By simultaneously analyzing microbiome profiles from the oral cavity, skin, and vagina, this method enables the classification of the tissue origin of forensic samples using artificial neural networks [104].
In addition to microbiome-based approaches, several classification methods have been developed that integrate multiple miRNA expression profiles detected by RT-qPCR. These methods apply continuous or probabilistic models to combine signals from multiple markers, thereby improving body-fluid discrimination compared with single-marker analyses [70, 71].
Collectively, these studies demonstrate that algorithmic approaches are becoming increasingly interwoven with spectroscopic, nucleic acid-based, and microbiome-derived data. Such data integration not only enhances classification performance but also represents a key methodological trend in the future development of saliva identification in forensic science.
This review summarized current knowledge on human saliva identification in forensic science and highlighted key methodological strengths, limitations, and practical considerations. Numerous saliva identification methods have been developed. Although these methods can be directly compared in terms of their detection sensitivity and specificity, these parameters alone do not determine their suitability for practical forensic applications. Other critical factors include compatibility with downstream DNA typing, cost-effectiveness and operational convenience, and clarity and explainability in court.
Crime-scene samples are rarely ideal, as heat exposure, substrate interference, environmental aging, mixed body fluids, and population structure (e.g., oral microbiome signatures) can complicate interpretation. Accordingly, validation should be population-aware, substrate- and aging-aware, and cognizant of co-analyzed modalities, including DNA typing, serology, and spectroscopy. No single approach currently satisfies all these criteria simultaneously. In practice, tiered workflows that comprise initial screening, confirmatory testing, and preservation of DNA for individualization are likely to provide the most balanced solution.
In forensic practice, it is important to understand the advantages and disadvantages of each saliva identification method before selecting an appropriate approach. The condition of the sample should also be considered. The development of new methods should prioritize not only detection sensitivity and specificity but also compatibility with other analyses, cost efficiency, and the ability to explain the method clearly for effective application in forensic investigations.
Ryota Komori: Conceptualization, Literature review, Writing – original draft. Tsutomu Nakagawa: Conceptualization, Writing – review & editing, Supervision. All authors have read and approved the final manuscript.
