Discussion

This systematic review provides an overview of the literature on amino acid profiles associated with advanced colorectal neoplasia. Overall, no consistent advanced neoplasia-specific amino acid signature was identified within or across matrices. Most amino acids were reported as non-differential in the majority of studies, and significant results were often conflicting, with the exception of tissue, where a consistent trend towards elevated amino acid levels in CRC tissue was observed. Diagnostic performance of individual amino acids ranged widely (AUC 0.28–0.91), with the highest values originating from a faecal-based study. A key observation is the substantial methodological heterogeneity across studies, in both analytical techniques and data processing methods, which likely contributes to the observed variability. In addition, studies addressing early lesion detection remain scarce, highlighting a clear gap in the literature regarding the biomarker potential of amino acids in early stage disease.

Most included studies described the diagnostic performance of individual amino acids as part of broader metabolomics analyses. Only four out of twenty studies describing diagnostic performance employed external validation; in two of these, fewer than ten CRC patients were included, while the remaining two involved 30 and 59 patients with CRC. None of these studies reported an AUC exceeding 0.80 [18, 23, 25, 27]. Internal validation was performed in three studies [14, 28, 29], whereas no form of validation was applied in the remaining 13, limiting the generalisability of findings and increasing the risk of overestimating diagnostic performance. Moreover, three studies included larger CRC cohorts (up to 250 patients) but described poor diagnostic performance (AUC < 0.66) [30–32]. In addition to sample size and validation, the analytical approach (targeted vs. untargeted) also affects the robustness of findings. Half of the studies applied an untargeted approach without subsequent analytical confirmation of the amino acids using a targeted method, which may lead to uncertainty regarding the reliability of the identified metabolites. Furthermore, in 15 studies evaluating diagnostic biomarker panels, the best-performing models combined amino acids with other metabolites or proteins, outperforming amino acid–only panels. This may be explained by the fact that such panels capture a broader spectrum of tumour-specific metabolic alterations, reflecting coordinated changes across interconnected pathways, including amino acid metabolism, lipid synthesis, and host–microbiome interactions [1].

When examining consistency within individual biological matrices, CRC tissue showed a notably consistent pattern of increased amino acid concentrations across multiple studies. This finding aligns with the Warburg effect, the metabolic shift of tumour cells towards aerobic glycolysis, in which increased anabolic demands are supported by amino acids acting as carbon and nitrogen sources for biosynthesis [33]. Across studies, glutamate, glycine, and phenylalanine were most frequently elevated. Increased glutamate levels are consistent with enhanced glutamine dependency, a well-established feature of tumour metabolism, whereby glutamine is converted to glutamate and subsequently fuels the tricarboxylic acid (TCA) cycle and biosynthetic processes [33, 34]. Moreover, glutamate is strongly related to 2-oxoglutarate, the latter being an essential metabolite in the TCA cycle [35]. Elevated glycine aligns with increased one-carbon metabolism, which supplies one-carbon units for nucleotide synthesis and methylation reactions essential for DNA replication and cell division [36]. In addition, glycine and glutamate, together with cysteine, support glutathione synthesis and the detoxification of reactive oxygen species in tumour cells [36]. Phenylalanine elevation may reflect increased protein synthesis and turnover associated with rapid tumour growth, in addition to its involvement as a precursor in nonessential amino acid synthesis [36]. Together, these alterations highlight metabolic adaptations that support tumour proliferation and provide mechanistic insight into the consistent amino acid enrichment observed in CRC tissue.

In plasma, leucine and valine were consistently downregulated in all studies that assessed them, suggesting a potential role for these branched-chain amino acids in tumour-associated metabolic reprogramming. Leucine and valine are key substrates in energy production and protein synthesis, but are also involved in cell signalling pathways that regulate growth and proliferation, such as mTOR activation [37]. Their depletion may reflect increased uptake by proliferating cancer cells or altered systemic metabolism. Interestingly, despite the physiological similarity between plasma and serum, the amino acids reported as differentially abundant showed limited overlap, with serum studies more often reporting non-significant findings. This aligns with previous observations that serum generally contains higher and more variable amino acid concentrations [38].

The observed discrepancies in amino acid alterations within and between matrices may be explained by several underlying factors. First, tumour-specific characteristics, such as disease stage, tumour location, and age of onset, have been associated with variation in amino acid profiles, possibly due to altered gut microbiota composition [2, 39–42]. In addition, differences in amino acid profiles may arise from inherent biological variation between matrices. These include matrix-specific metabolic activity, differences in host–microbiota interactions, and variable amino acid stability or degradation during sample collection and processing [43–45]. Furthermore, although lifestyle-related factors, particularly diet, can significantly affect amino acid concentrations across sample types, most included studies rarely accounted for lifestyle variables [4, 46, 47]. While tumour-related and matrix-specific biological factors likely contribute to genuine variability in amino acid profiles, the inconsistent consideration of diet and variation in sample processing across studies suggests that a substantial proportion of the observed discrepancies may be methodological rather than biological in origin. Future studies should incorporate detailed dietary assessments and standardised sampling conditions (e.g. fasting versus non-fasting) to reduce this source of confounding.

Variation in analytical methods across studies likely represents an additional contributor to the inconsistencies observed in amino acid profiles. Although this review focused on the direction of amino acid alterations rather than absolute concentrations, differences in analytical sensitivity and dynamic range may still influence whether subtle group-level changes are detected. NMR, HPLC, and various mass spectrometry (MS) platforms were used, each differing in sensitivity, selectivity, and quantification accuracy. Although one study reported strong correlations between NMR and LC–MS/MS for most amino acids, systematic differences in absolute concentrations were observed, and group-level differences were not consistently detected across both methods [48]. HPLC remains a commonly used technique in amino acid analysis, yet coupling it to MS can substantially enhance sensitivity, especially in complex matrices such as faeces [49]. Even within MS-based approaches, methodological variation (e.g., GC–MS vs. LC–MS or CE–MS) may influence which amino acids are reliably quantified or detected [50, 51]. Variability in targeted versus untargeted approaches may further influence detection sensitivity and coverage [52], however, discrepancies in amino acid profiles were observed even among studies applying comparable analytical strategies. In addition, studies differed in their statistical approaches, including the use of univariate versus multivariate methods, correction for multiple testing, and criteria for reporting significance. Together, these methodological differences may contribute to the lack of consistency in amino acid profiles across studies.

A key strength of this review is its comprehensive scope, combining an extensive literature search with a structured comparison of amino acid profiles across multiple biological matrices. By evaluating findings from faeces, urine, serum, plasma, tissue and saliva side by side, this review provides a broad overview of matrix-specific trends and inconsistencies. In addition, diagnostic potential was assessed where applicable, offering clinically relevant insight beyond descriptive profiling.

Limitations

This review has several limitations. First, a substantial methodological heterogeneity across studies, including differences in analytical techniques, statistical approaches, and study design, precluded formal meta-analysis and limited comparability. In addition, several included studies had small sample sizes, hampering power of their findings. Potential publication bias should also be considered. Untargeted metabolomics studies frequently reported only statistically significant results, while non-significant amino acids were often omitted. This selective reporting may have led to an overestimation of differential findings. To mitigate this, targeted and untargeted studies were analysed separately. Moreover, unaccounted heterogeneity related to pre-analytical factors, microbiota composition, and tumour characteristics may also have contributed to bias.

Future perspectives

Future research should prioritise standardisation of analytical protocols to improve comparability between studies, as well as focus on early disease stages, particularly advanced adenomas and advanced serrated polyps. Identifying biomarkers that detect precancerous lesions remains critical for preventing their progression to CRC by endoscopic polypectomy. Importantly, rigorous validation strategies, including external validation, are essential to assess the true clinical utility of candidate biomarkers. Given that diagnostic performance often declines upon external validation, larger, well-characterised cohorts and harmonised methodologies will be key to advancing the field.