**4. Composite biomarkers in non-invasive sampling: what is known and what could be useful in the future?**

The complexity and dynamics of asthma drive the need to establish distinct disease phenotypes and endotypes. There are several different triggers in asthma, with various pathophysiological pathways in parallel resulting in clinical expression that may be rather similar. Therefore, repeated multiscale, multidimensional measurements may be needed to capture this com‐ plexity, which may yield more useful information than single or even panels of combined biomarkers [60]. In this view, molecular composite signatures may be obtained by highthroughput "omics" technologies, which are increasingly standardised. Several large-scale studies of the genome, transcriptome, proteome or metabolome have produced an enormous amount of data and it is pivotal to follow the available guidelines in order to avoid false discoveries [60]. The composite high-dimensional signatures or fingerprints are based on pattern recognition underlying complex non-linear biology systems. Some evidence that this approach may be successful in asthma has already emerged concerning differential diagnosis [61]. Regarding non-invasive, direct assessment of exhaled breath, it is interesting to note that while many problems arise in specific molecular biomarkers validation, recent studies have shown encouraging results with the application of metabolomics strategies to study exhaled biomarkers [60, 62, 63].

adults, and in patients with aspirin-exacerbated respiratory disease [55, 56]. It has been associated with the degree of airflow limitation and acute exacerbations [55, 57]. Although inhaled corticosteroids are the most effective treatment for asthma, they do not alter LTE4 excretion. Urinary LTE4 levels have been suggested as potential predictors of better response to anti-LT therapy compared to other therapeutic approaches, though further studies are

As LTs, *prostaglandins* (PG) are the end products of arachidonic metabolism. PGD2 results from cyclooxygenase pathway and is excreted in urine after being metabolised to 9α,11β-PGF2. Increased urinary excretion occurs in patients with asthma and after challenge tests, and a

*Bromotyrosine* is another molecule with possible interest in asthma. It is generated from protein oxidation by eosinophils. The oxidised amino acid is stable and excreted in urine. Urinary bromotyrosine levels are higher in patients with asthma and have been associated with asthma control and lung function, predicting exacerbations [59]. Its levels have been shown to reduce during inhaled corticosteroid therapy. High urinary bromotyrosine levels could predict a favorable clinical response to inhaled corticosteroid therapy, especially in combination with

Though urinary biomarkers may become useful tools, many require specialised equipment and their measurement is not fully validated or standardised. There is a current need for normalisation standards and assessment of intra and inter-individual variation to select the potentially useful biomarkers. It is also important to address urine dilution when reporting

Airway imaging biomarkers are also emerging, offering the potential of adding complemen‐ tary information, namely on small airways function and remodeling. High-resolution com‐ puted tomography (HRCT) images are used to measure airway narrowing, wall thickening, air trapping and ventilation inhomogeneity [27]. The first two measures have been correlated with lung function and asthma severity. Increased parenchymal lucency has also been associated with severe asthma exacerbations, lung function and neutrophilic inflammation. HRCT is easily performed though it requires that lungs are scanned at a standard volume for validity and reproducibility. The risk of exposure to significant ionising radiation needs to be

**4. Composite biomarkers in non-invasive sampling: what is known and**

The complexity and dynamics of asthma drive the need to establish distinct disease phenotypes and endotypes. There are several different triggers in asthma, with various pathophysiological pathways in parallel resulting in clinical expression that may be rather similar. Therefore,

needed, including other biomarkers, to predict individual responses.

high FeNO values [59]. These results warrant further developments.

considered and normal ranges have not been established.

**what could be useful in the future?**

quantitative absolute results.

**3.5. Airway imaging biomarkers**

negative association has been found with lung function [58].

24 Asthma - From Childhood Asthma to ACOS Phenotypes

Among "omics" systems biology, metabolomics is considered the one that comes closest to phenotype expression. It involves the identification and quantification of small molecular weight metabolites. Real-time metabolomics measurements are already feasible for several clinical applications with electronic noses (eNoses). These handheld, portable devices can capture various combinations of VOCs in exhaled breath, with nanosensors arrays. The nanosensors are based on conducting polymers, metal oxide, metal oxide field effect transis‐ tors, surface or bulk acoustic waves, optical sensors, colorimetric sensors, ion mobility spectrometry, infrared spectroscopy, gold nanoparticles and gas-chromatography (GC) coupled with mass spectrometry (MS) or flame ionisation detection [60, 64]. The pattern recognition algorithms using various eNose sensor systems indicate fingerprints of exhaled VOCs, called breathprints, which have shown to discriminate patients with asthma from healthy subjects and COPD with accuracies between 80% and 100% [65]. Breathprints have also been studied to phenotype asthma. Recent studies indicate that eosinophilic and noneosinophilic asthma can be distinguished when using a composite eNose platform. Breath analysis by eNose could also predict the response to corticosteroids with greater accuracy than sputum eosinophils or FeNO [66]. These data suggest that composite signatures of breath analysis could be used for assessment and monitoring of airway inflammation. Important methodologic issues of technique optimisation and standardisation deserve deeper analysis, from breath sampling, to modulating factors including comorbidities and incompatibility between eNoses. These should enable external validation to determine possible diseasespecific breathprints with clinical applicability.

Besides the analysis of exhaled breath vapour with GC-MS and eNoses, the novel metabolo‐ mics approach has also been applied to EBC. It has been shown to enable characterisation of metabolic compounds in even small EBC volumes, using high-resolution proton nuclear magnetic resonance (NMR) or MS. This has proved capable of discriminating healthy indi‐ viduals from those with asthma [62, 63, 67, 68]. It could also discriminate between severe and non-severe asthma [63], supporting the hypothesis that severe asthma has specific metabolic features.

Interestingly, the metabolomics analysis of urine also discriminated healthy individuals from those with asthma [69], and could distinguish patients with stable asthma from those with acute exacerbations based on profiles [69, 70]. Metabolomics analysis of urine samples has also been recently suggested as a useful clinical tool to differentiate asthma from COPD [71].

Pinkerton et al. [72] demonstrated for the first time that differences between healthy controls and asthma patients could be detected via micro-RNA (miRNA) expression in EBC, and suggest that different types of inflammation may have unique miRNA signatures. These small non-coding RNAs are known to be important in the post-transcriptional regulation of inflam‐ mation, thus opening a new research field using non-invasive direct air sampling.

Proteomics has also recently been applied to EBC. Liquid chromatography (LC)-MS has been used to separate and detect proteolytic peptides present in EBC with differentiating profiles based on asthma status [73]. However, this preliminary study faced several problems such as insufficient sample volume, possible salivary contamination and difficulties in peptides identification due to their low concentration.

Besides allowing an overview of molecular signatures, the "omics" approach may potentially lead to new knowledge regarding asthma pathophysiology, due to its untargeted, hypothesisgenerating approach. All biomedical researchers are facing not only the opportunities but also the challenges in accessing, managing, analysing and integrating diverse data sets that are larger, more diverse and more complex than ever before, and that exceed the abilities of current management and analysis approaches [60, 74]. Composite biomarkers research such as that coming from molecular profiling assays including various "omics" is a live example that needs to be critically interpreted and cautiously validated to yield truly significant advances in personalised medicine.
