Breathomics is an emerging science to diagnose diseases by analyzing volatile organic compounds (VOCs) produced by changes in metabolic processes caused by disease [1].
The electronic nose (E-nose) and gas chromatography-mass spectrometry (GC-MS) are two methods to analyze these VOCs. In contrast to the E-nose, GC-MS allows us to explore possible biological pathways and identify specific VOCs associated with the pathological changes of the diseases.
Currently, the E-nose has been implemented in the diagnosis of lung cancer [4], breast cancer [5], colorectal cancer [6], ovarian cancer [7], gastric cancer [8], head-and-neck cancer [9], chronic obstructive lung disease (COPD) [10], interstitial lung disease [11], liver cirrhosis [12], ventilator-associated pneumonia [13], and Coronavirus Disease 2019 (COVID-19) [14].
In artificial intelligence (AI), the development of electronic nose systems is an emerging science that can provide real-time analysis and assist clinical decisions. There are two major types of sensors: (1) nanomaterial-based sensors, including single-walled carbon nanotubes (CNTs), monolayer capped metal nanoparticle (MCNP) films and metal oxide (MO) sensors, and (2) electroacoustic sensors that include quartz microbalance (QMB) and surface acoustic wave (SAW) sensors [15].
The current knowledge gap on the application of E-noses to clinical diagnosis remains uncertain. Due to the advancement of material sciences, many types of E-nose sensors have been developed in recent years [16]. Although many types of sensors have been designed to detect more diseases in recent years, E-noses have not yet been applied in clinical practice
Preventive diagnosis
VOCs are organic chemicals that have a high vapor pressure at room temperature, i.e., low boiling points roughly in the range of 50 to 250 °C [15–17]. Most VOCs have unique odors [12, 18, 19]. There is a consensus between eastern and western traditional diagnosis of methodologies that smelling the breath from the upper respiration tract is an effective way for clarifying health conditions both in Hippocrates’s Book of Prognostic written in 400 B.C. [20] and Bian Que’s book of “The Yellow Emperor’s Canon of 81 Difficult Issues” written in ca. 200 B.C. [21].
The nature of these diseases related special smells is caused by different VOCs, which is demonstrated by the nobel prize laureate that Linus Pauling’s pioneer work on the breath analysis in 1971 using gas–liquid partition chromatography [22]. In the past 50 years, thousands of volatile biomarkers, i.e., VOCs, associated with various diseases or lesions, have been identified and classified from multi-body sources (Fig. 2) [23].
It boosts the development of this new era, volatolomics, such as human exhalation [18], skin emanations [24], urine headspace [25], blood [26], and feces [27], towards good-efficient, high-accuracy non-invasive, and painless disease diagnosis/screening.
Before the volatolomics diagnosis could be accepted as a new clinical golden standard, two major concerns should be well answered. One is what the solid relationship, between the specific set of volatolome and the aimed disease is, viz., what the volatolomics (bio-/chemical markers) of a special disease is.
The other one is how the volatolome can be accurately and specifically detected and recognized in the real environment fulfilled with confounding factors. According to these two concerns, what is the philosophy of research methodology to answer these questions in such an interdisciplinary research field? Some critical review articles had been published [28–32], in which these concerns were partly considered.
However, a systematical summary and elaboration is still necessary. In this review article, the research methodology of volatolomics in healthcare is critically thought and given out. Then, the sets of volatolome according to specific diseases through different body sources, and the analytical instruments for their identifications are systematically
summarized. Thirdly, the advanced electronic nose (E-nose) and photonic nose (P-nose) technologies for VOCs detection are well introduced. The existed obstacles and future perspectives are deeply thought and discussed. This article could give a good guidance to researchers in this interdisciplinary field, not only understanding the cutting-edge detection technologies for doctors (medicinal background), but also referring to clarify the choice of aimed VOCs during the sensor research for chemists, materials scientists, electronics engineers, etc.
2 The research methodology of volatolomics in healthcare
The emitted volatolome are complicated mixtures with confounding factors, e.g., food, smoking, alcohol drinking, environmental pollutions, etc., no matter which sources of the human body they come from. Therefore, to ensure the successful application in healthcare, a universal methodology is required to make all research results comparable. A brief introduction is depicted as a flow chart in Fig. 3.
First, volatolomics, the accurate bio-/chemical markers associated with the specific diseases, needs to be identified by analytical instrumental approaches, such as gas chromatography- mass spectrometry (GC-MS) [33–36], gas chromatography time- of-flight mass spectrometry (GC-ToF-MS) [37–39], etc., which is the building block of this new era. In addition, it is worth mentioning that the analytical instrumental approaches are still the most accurate way to analyze volatolome mixtures, although they are expensive, time-assuming, and high skill threshold. Nowadays, with the development of data mining technology, e.g., association analysis, clusters analysis, classification, and regression [40], accurate results can be obtained efficiently, facing big and complicated clinical data when all of them use the same sampling method.
Second, the sensing technology, E-nose [41] and P-nose [42], including sensitive materials, fabrications, recognition algorithm, etc., need to be developed according to the identified volatolome markers. In detail, the interaction between the volatolome and sensitive materials, needs to be well designed and verified according to different conditions. For example, in the disposable application such as P-nose, one-lock one-key strategy is widely used [28].
The sensor devices of fabrication process also need to be well considered. For example, the uniformity, scalability, and reproducibility of fabrication technology should be considered at the starting stage of research to narrow the big gap between laboratory and industry. The recognition algorithm is an effective tool to analyze data. The core includes feature’s extraction and analysis (cluster, separation, etc.) [43]. Good research on recognition algorithms not only can enhance the performance of
E-nose and P-nose, but also can help the researchers to understand the sensing mechanism through the extracted features. In one word, the development of sensing technology is based on the volatolomics, and then, verified by the volatolomics.
Third, the developed sensors need to be validated by strict clinical trials with confounding factors. In this stage, the recognition algorithm is further trained by the real samples from the patients. The larger number of the tested patients, the more precise algorithm is, which is the precondition of the success in the blind test. After that, volatolomics might be considered as a candidate of golden standard of diagnostics. Till that, the volatolomics can start to apply the admission from food and drug administration (FDA), and might be authorized in the future.
The volatolomics of specific diseases from different body sources
The volatile organic compounds in breath and their associated diseases
Human breath is a gas exchange process mainly for inhaling O2 and exhaling CO2. Accompanying this process, large amounts of volatile metabolites are produced in some normal and abnormal metabolic biochemical pathways. Currently, various kinds of VOCs have been found in breath samples, thus, making it the most examined VOC source. The main pathways for adding VOCs to breath include but are not limited as:
• From the viewpoint of thermodynamics, there is a VOCs equilibrium distribution between “fat–blood–breath”. The concentrations of VOCs in each part follow the fat–blood (λf:b) and blood–air (λb:a) equilibrium partition coefficients. Most cancer related VOCs are transported from different body sources to breathe in this way (Path ①) [44–58].
• The cell/tissue lesion in the mouth, on the inner surface of the alimentary canal and stomach, i.e., their headspace connects the respiratory tract, can directly emit VOCs to the breath (Path ②) [33, 45, 59–71].
• Infections caused odorous metabolites in the surrounded micro-environment, e.g., oral interstitial, helicobacter pylori, lung tuberculosis infections, etc., can directly emit VOCs to the breath (Path ③) [72–78].
Most diseases related VOCs, as shown in Fig. 4, and summarized in Table 1, are emitted along one or combined paths. The percentage’s sum of alkane and alkene are higher in Alzheimer, lung cancer, breast cancer, colorectal cancer, Parkinson, head-and-neck cancer, and renal disease patients’ breath, than those in others.
This means such VOCs transport mainly, but not limited, along the Path ①, viz., “fat–blood–breath” routine. The percentage’s sum of aldehyde, ketone, nitrile, amine, ester, alcohol, acid, i.e., polar VOCs, are higher in the gastrointestinal disease, gastric cancer, pulmonary arterial, and asthma patients’ breath, than those in others. These diseases can directly emit their metabolites into the breath, i.e., they are mainly, but not limited, along the Path ②. The halitosis, chronic obstructive pulmonary disease (COPD), and tuberculosis are mainly along Path ③ and other combined paths.
Volatile organic compounds in blood and their associated diseases
Blood circulates in the whole body, exchange, and transport substance in most bio-chemical process, reflect the real-time nutritional, metabolic, and immune status. Theoretically, blood volatolomics could reveal full information of all diseases in the human body. The blood-related VOCs metabolic processes can be classified into two categories as follow:
• Damage to the cells by reactive oxygen species through direct oxidative stress processes generates a major source of VOCs that partially exchange into blood (Path ①) [32].
• Exogenous VOCs, e.g., food, air pollution, and smoking, cause indirect oxidative stress. Those VOCs leak into the cytoplasm, then, attach to organs or organelles. The subsequent peroxidative damage to proteins, PUFAs and DNA produce VOCs that partially exchange into blood (Path ②) [30, 79].
Till now, many studies only examined the VOC profiles through in-vitro cell experiments that are greatly different from the real conditions surrounding tissue and blood vessels. The aforementioned thermodynamics model about VOCs equilibrium distribution between “fat–blood” decides the VOCs input by the fat–blood coefficient (λf:b).
The exchange processes of VOCs from blood to breath, urine, feces, and other things dominate the output. Therefore, the input and output of the blood VOCs decide the compositions and concentrations. However, in fact, the studies of volatolomics are not well concerned compared with the studies on biomarkers in blood.
Very few researches show the VOCs in liver cancer, lung cancer, and hepatic cancer patients’ blood as shown in Fig. 5 and summarized in Table 2. The less kinds of VOCs than that in breath and urine seem lost large amounts of information. The possible reasons might be:
• Fast solidification of blood makes hard sampling and examinations, although the study towards the blood VOCs is very important to understand the VOCs’s generation and transport mechanism in the human body.
• The blood is not stable for storage due to the existence of erythrocyte, leucocyte, enzyme, bacterial, etc. These “dirties” can consume and/or emit VOCs in the blood, which makes the blood VOCs analysis complicated.
• Relatively low concentration of VOCs exists in the blood without any preconcentrating process compared with VOCs in urine, which makes concentrations of VOCs, sometimes, lower than the limit of detection of spectrometry instruments.
Therefore, the single volatolomics associated with blood in diagnostics still needs to be further studied.
The volatile organic compounds in excreta and their associated diseases
Human VOCs are present in a variety of excreta, especially urine and feces. There are two main advantages for VOCs in excreta. First, the VOCs can be directly released to the air along the intestine and urethra. Second, the excreta can be a good media for capturing and taking VOCs out of the body. The main pathways for emitting VOCs to excreta include but not limited as:
• The similar emission path as Path ① is mentioned in Section The cancer-related VOCs are transported from different body sources to the intestine and urethra according to the “fat–blood–air” and “fat–blood–water” equilibrium distribution [80].
• The cell/tissue lesion on the inner surface of the intestine and urethra, i.e., their headspaces connect the environment, can directly emit VOCs to the air or water (Path ②) [81–85].
• Infections caused odorous metabolites in the surrounded micro-environment, e.g., urinary tract inflammation, enteritis, etc.,can directly emit VOCs to the intestine and urethra, captured by the excreta (Path ③) [86, 87].
• The disease caused disorder of symbiotic bacteria (mainly in feces), e.g., Escherichia coli, can interfere with the bio-chemical reaction of food decomposition and absorption, which changes the compositions and concentrations of the VOCs (Path ④) [88–92].
• The VOCs emitted by tumor cells and/or lesions on the digestive tract or the organs nearby the digestive tract can be mixed with food and finally excreted in the feces along with the digestive process (Path ⑤) [93–95].
The VOCs in urine and feces contain abundant physiological information in volatolomics as shown in Fig. 6, and summarized in Tables 3 and 4. VOCs, mostly polar molecules such as acid, alcohol, ketone, and aldehyde, in urine, are directly linked to infectious diseases, cancers, and particular disorders. No alkane has been found in the urine, which is decided by the poor solubility of non-polar molecules in it. In the feces, ester, alcohol, ketone, and acid are mostly found. Interestingly, the volatolomes of the same disease in breath, urine, and feces are totally different, which means the distribution and emission path of VOCs are greatly different in different body sources.
Urine and feces volatolomics as the promising diagnostic option can be summarized as follow aspects:
• Urine samples are easier to be sampled and stored than invasively collected blood and non-invasively collected breath. Furthermore, when stored under the right conditions, urinary VOC samples are stable for long periods.
• Higher concentration of VOCs is obtained by the “pre- concentration” process in the kidney and semi-solid extraction process in the digestive tract. In other words, the key “VOCs signature” is “recognized and amplified” with increasing signal-to- noise ratios.
• Urine is much less affected by diet or the other commensal of bacteria in the gut than feces or breath.
The volatile organic compounds on skin and their associated diseases
Skin is one of the largest organs for the human beings. The skin has a layered structure to perform as a barrier and achieve the matter exchange between body and air. Therefore, VOCs from the skin originate from four major sources：
• The gland secretion, which is related to the matter exchange process (Path ①) [184].
• The skin diseased related VOCs directly are emitted to air, and absorbed on the skin (Path ②) [185–189].
• The metabolites of skin microbiota that is barred out of the body but stayed on the skin surface (Path ③) [190–193].
• The exogenous VOCs, e.g., food, air pollution, and smoking, which are absorbed on the skin surface (Path ④) [194].
Eccrine excretion (sweat), similar to urine, usually contains 98% water, with the rest being various organic and inorganic compounds (sodium chloride, lactate, and urea), which is mainly transferred by osmosis. Extracellular fluid is the origin of eccrine secretion, and thus reflects blood plasma chemistry. However, the skin VOCs caused by eccrine excretion are very different from the urinary VOCs as shown in Fig. 7 and summarized in Table 5.
For example, the alkanes, rare to be found in urine, can be detected in melanoma skin samples. The profiles of tuberculosis volatolome are identified from clinical samples. There is also a difference between breath and skin volatolome. Very interesting, for aging people, the 2-nonenal is clearly confirmed as the chemical markers comparing the volatolome emitted by people elder and younger than 40 years old [183]. According to the bacteria infection, acids are very frequently found in the volatolome, which may be related to fat oxidation and other pathways.
More than 500 VOCs have been identified from human skin extracts, however, the identifications of volatile markers towards specific diseases in/on the skin are still challenged due to more confounding factors on body odor, such as food, environment, and body clean, which requires a more advanced algorithm to assist the analysis of spectrometry results.
The latest efforts on diagnosis towards tuberculosis through detecting the skin VOCs by E-nose had been achieved in India and southern Africa [24], which might be an effective demonstration for developing new wearables and skin electronics, based on the skin volatolome, in the future.
The detection of volatile organic compounds by E-nose and P-nose
The individual sensor device is the basic functional element, and thus determines the overall performances of artificial olfaction as a whole (Fig. 8). The basic gas sensors for breath analysis are classified into three categories according to the transducer (i.e., mass, electrical and optical) and are discussed below with selected examples (Fig. 8). Most of the E-noses are constructed by the same type of gas sensors with distinct performances (homo gas sensor arrays or homo-transducer arrays).
Recently, E-nose consists of different types of sensors that have been developed (hybrid gas sensor arrays or multi-transducer arrays) [195, 196], which can reduce correlations between the responses of different sensor types, and thus improve the power of data analysis [197–199].
The quartz crystal microbalance (QCM) [200, 201] and surface acoustic wave (SAW) [202, 203] sensors are two common types of mass transduced gas sensor arrays applied in breath analysis. As for optical transduced gas sensors, colorimeter [204, 205] and nondispersive infrared (NDIR, for CO2 monitoring) [206] types are used.
Different from optical or mass transduced sensors, electrically-transduced gas sensors normally required semiconductive/conductive ability or other sensitive electrical properties of the active materials (e.g., capacitance (C), inductance (L), resistance (R), and work function (ϕ)) [207, 208]. Various electrically-transduced gas sensors including chemireistors [41],field-effect transistors (FETs) [209], chemical diodes [207, 210], chemicapacitors [211], and electrochemical sensors [197, 212] are excellent candidates for the construction of E-nose [43, 213].
For mass-transduced E-nose for the detection of VOCs, a typical example is the use of SAW sensor arrays, in which a biomimetic olfactory receptor-based biosensor with better performances was reported by improving the immobilization efficiency of molecular detectors for breath analysis (Fig. 9) [202]. The specific olfactory receptors (ODR-10) were used to functionalize the sensitive area of the SAW chip with a self- assembled monolayers (SAMs) of 16-mercaptohexadecanoic acid (MHDA).
The responses to the various VOCs were recorded by monitoring the mass loading affected the responance frequency shifts of SAW. The functionalized SAW showed the ultrasensitive detection (LOD = 1.2 × 10−11 mM) with high selectivity compared to butanone and 2,3-pentanedione [202].
Chemiresistors are the extensively studied types of electrically- transduced gas sensors for E-nose. For example, gold nanoparticle (GNP) based E-noses was reported in the successful diagnosis of lung cancer using breath samples from real patients (Fig. 10(a)) [41]. Firstly, the sensors were selected and trained by simulated vapors of healthy and cancerous breath according to GC-MS results (Fig. 10(b)).
Then the selected sensor arrays were exposed to the breath of normal healthy and cancerous patients. Principal component analysis (PCA), a commonly used statistical procedure to reduce a set of possibly correlated variables into a set of values of linearly uncorrelated variables via an orthogonal transformation [43, 213–215], was employed in the post-treatment of the sensing results.
The complete separation of the output PCA results indicated that such sensor arrays could discriminate between the different smells of healthy people and patients with lung cancer (Fig. 10(c)). Representative works of E-nose based on other chemiresistive materials such as chemically modified carbon nanotubes (CNTs) [216, 217], metal oxides [43, 218], surface- modified Si nanowires [219], and polymer-carbon hybrids [220, 221] were also successfully applied to disease diagnosis with high accuracy.
In addition, the newly emerging sensing materials such as crystalline porous materials (e.g., metal-organic framework/porous coordination polymers [222–226], black phosphorus [227–230], covalent-organic frameworks (COFs) [231, 232], and Mxenes [233–235]) and their hybrid materials/thin films [236, 237] based chemiresistors with low temperature sensitivity and high selectivity can also be applied in E-nose in the near future.
In most cases of E-nose, studies have concentrated on a single array technology and single transduction technology, such as electric, mass, and optic. Most e-noses use cross-interactive sensor arrays that react to the VOCs on the sensitive materials’ surface attaching with adsorption, desorption and/or reversible reaction, etc.
Then the specific responses between the VOCs and sensors array are recorded and transformed into readable digital values which can achieve the recognition and detection based on statistical models or machine algorithm [238–242]. For the hybrid gas sensor arrays (multi-transducer arrays), different transduced techniques are used to lower correlations between the responses of the different sensor types, that greatly improve the accuracy of data analysis and pattern recognition [197, 243].
Therefore, a hybrid gas sensor array endowing with carefully selected sensing materials (receptor) and multivariable transducers is highly promising for the high-performance recognition of breath VOCs [196, 198, 244–246]. By using the same sensing materials (receptor) while distinct transducer, QCM, and chemiresistive platforms were used to investigate the selectivity of one pristine and three surface-modified single-walled carbon nanotubes (SWCNTs) to VOCs (Fig.11(a)) [199].
As for the electrically- transduced sensor, the VOC molecules take longer to intercalate between the junctions of 2 nanotubes, thus delaying the increase in resistance (Fig. 11(b)). The signal responses of SWCNT-coated QCMs and chemiresistors to butanol and butyl acetate are shown in Figs. 11(c) and 11(d) [199].
The two cases showed the QCMs reached equilibrium faster than the chemiresistors did. Such additional mechanism involved in changes of resistance resulted in that, even with the same active materials, the response time, linearity, and relative sensitivities were all different between QCM and chemiresistor arrays. The diverse response patterns on two different sensor array transducers enable the formation of a highly selective hybrid gas sensor array.
The colorimetric sensor array, a kind of P-nose, represents a facile and visible optical-transduced detecting approach, the color patterns of which change upon exposure to VOCs [247]. The typical system of colorimetric sensing is shown in Fig. 12(a). The core parts include chamber, digital cameras, light source, gas inject controller, pump, and some sensors, which has great potential to further minimize accompanying with the development of the electronics industry [248].
Such systems, normally, achieve breath diagnosis with higher accuracy and specificity, due to the probes loaded on the sensing materials using one-lock one key strategy. Changjun Hou et al., have finished a series of works on lung cancer diagnosis-based colorimetric sensing technology [19, 248–251]. Figure 12(b) shows that the colors of the sensing matrix are changed after exposure to the breath of a lung cancer patient, in which the differences can be further magnified by the advanced algorithm. In addition, the sensor array also showed good anti- humidity properties, which avoids the common problem of humidity interference in practical applications. The main limitation of such colorimetric sensor arrays is its irreversibility.
reference link : https://www.mdpi.com/2079-6374/11/11/469/htm
https://doi.org/10.1007/s12274-022-4459-3
