Metabolic insights into tumor pathogenesis: Unveiling pan‐cancer metabolism and the potential of untargeted metabolomics

Abstract

Metabolic dysregulation is a hallmark of cancer, underpinning diverse aggressive behaviors such as uncontrolled proliferation, immune evasion, and metastasis. Despite the potential of tumor metabolites as biomarkers, their utility has been hampered by metabolic heterogeneity. Exploring cancer metabolism aims to discern shared metabolic pathways and have a better understanding the metabolic heterogeneity of tumors. This approach offers a holistic view of cancer metabolism, facilitating the identification of multicancer‐relevant metabolic targets and the development of more broadly effective therapeutics. In this review, we present a comprehensive overview of the current landscape of cancer metabolism and its prospective applications in cancer diagnosis and prognosis. We delineate common metabolic aberrations observed across a spectrum of cancer types and elucidate the unique metabolic signatures characterizing the six leading causes of cancer‐related mortality. Furthermore, we survey the utilization of untargeted metabolomics and single‐cell technologies in cancer screening, diagnosis, and prognosis, while also spotlighting available data resources for pan‐cancer metabolomics analyses. Throughout this discussion, we tackle prevailing research challenges and propose strategies aimed at enhancing cancer management. Our objective is to furnish valuable insights that can inform and guide future research endeavors in the dynamic realm of cancer metabolism.

INTRODUCTION

Noncommunicable diseases (NCDs) have emerged as a significant source of worldwide mortality, with cancer being a prominent contributor. Cancer is a complex disease that is characterized by uncontrolled cell growth and proliferation, and its metabolic rewiring is a key factor in cancer progression. Since the discovery of the Warburg effect nearly a century ago, in which tumor cells significantly upregulate glycolytic pathway even in the presence of oxygen, metabolic alterations in cancer have remained a topic of interest for researchers (Figure ). Cancer cells must reprogram their metabolism to meet the demands of uncontrolled replication and adapt to the microenvironment. Understanding the metabolic changes in cancer cells offers insight into tumorigenesis and provides a potential source of new biomarkers. Pan-cancer metabolic profiling can provide a comprehensive understanding of different tumor types as well as different subtypes of the same tumor and improve metabolism-based pan-cancer analysis to identify common targets and biomarkers for multiple cancer types.

[IMAGE OMITTED. SEE PDF]

The purpose of this review is to provide an overview of current knowledge of metabolism in cancer biology and opportunities for metabolomics in cancer diagnosis and prognosis. We begin by summarizing common and specific metabolic changes in six leading causes of cancer death: lung cancer, liver cancer, gastric cancer, colorectal cancer, esophageal cancer, and breast cancer. We then review current applications of metabolites in cancer diagnosis and prognosis and list commonly used public resources for metabolomics analysis. In addition, we discussed the application of single-cell technologies that accelerates cancer metabolism research. Finally, we discuss limitations and opportunities for current metabolomic research. Since metabolomics analysis tools have been extensively covered in other excellent reviews, readers can refer to them for more in-depth discussion (Figure ).

[IMAGE OMITTED. SEE PDF]

Metabolomics is the science of quantitatively describing the patterns of biological endogenous metabolites in response to changes in internal and external factors, which can directly reflect the terminal and phenotypic information of living organisms. In this review, we introduce the application of metabolomics in cancer and the prospect of combining metabolomics with other histologies, aiming to provide a theoretical basis for the exploration of highly sensitive and specific tumor markers, which can facilitate the early detection and treatment of cancer, the identification of potential therapeutic targets, as well as the improvement of patient survival.

METABOLISM IN TUMORS

In this section, we focus on summarizing the changes in several important metabolic pathways in tumor cells and metabolic changes specific to six major cancers (lung, liver, gastric, colorectal, esophageal, and breast cancer). Through in-depth understanding of the metabolic regulatory mechanisms of diseases, we will discover biomarkers for disease diagnosis, identify drug targets suitable for disease treatment, and realize personalized and precise medical treatment.

Glucose metabolism in tumors

As we all know, energy is the foundation of life. To sustain their metabolic machinery during senescence or while they are dividing for anabolism, all cells need energy. In contrast to normal cells, cancer cells are constantly dividing and require significant metabolic changes to meet the demands of unchecked growth. Glucose is the major source of cellular energy. In the presence of oxygen, glucose is oxidized through the tricarboxylic acid (TCA) cycle and oxidative phosphoric acid, eventually producing carbon dioxide and water and releasing large amounts of energy. Cells ferment or anaerobic glycolyze in the absence of oxygen, and pyruvate is converted to lactate or alcohol, at which point less energy is generated. However, in the 1920s, Warburg discovered enhanced glucose uptake and lactic acid production in tumor cells compared to normal cells, even in the presence of oxygen. This phenomenon termed as Warburg effect or aerobic glycolysis. Although glycolysis produces a small amount of Adenosine triphosphate (ATP), It has been discovered that the Warburg effect causes a similar quantity of ATP to be produced over the same period of time since the flow of glucose to lactose is 100 times faster than that through the TCA cycle. In addition, the intermediates productions of the Warburg effect provide a carbon source for nucleotide metabolism, lipid metabolism, and amino acid metabolism, maintaining the proliferation and differentiation of tumor cells.

For example, glucose can be transferred to the serine synthesis pathway via the glycolytic intermediate 3-phosphoglycerate (3PG) and eventually converts serine to glycine. Serine and glycine are intermediates in the pathway for the synthesis of other amino acids as well as lipids and nucleic acids. In the pentose phosphate pathway (PPP), glucose 6-phosphate (G6P) is partially oxidized to produce nicotinamide adenine dinucleotide phosphate (NADPH) and ribose-5-phosphate (R5P). R5P is a structural component of nucleotides. To facilitate the reductive production of fatty acids and nucleotides, NADPH is a crucial anabolic reducing agent. Additionally, NADPH keeps cells alive in oxidative stress situations brought on by mitochondrial dysfunction or metabolically active cells. Glycerol-3-phosphate (G3P) converted from dihydroxyacetone phosphate (DHAP) is the raw material for phospholipid synthesis. Fructose 6-phosphate, produced in glycolysis and enter the hexosamine biosynthesis pathway (HBP), is suggested as an energy sensor of the cell. In addition, studies have proved that lactic acid produced by glycolysis can not only be converted into glucose through gluconeogenesis but also be absorbed by some cells to fuel the TCA cycle.

Glucose enters the cell via glucose transporter protein (GLUT) and is eventually produced as pyruvate through glycolysis. During this process, oncogenes c-MYC, KRAS, and YAP upregulate GLUT1 expression in cancer cells. It has been shown that protein hypoxia-inducible factor-1α (HIF-1α) can also upregulate GULT1 and promote glycolysis. The phosphoinositide-3 kinase (PI3K), protein kinase B (AKT), and mammalian target of rapamycin (mTOR) play an important role in cell metabolism. Numerous cellular stimuli, including low levels of nutrition, oxygen supply, pH, growth hormones, and receptor tyrosine kinases, have been demonstrated to activate these signaling pathways. AKT is a serine-threonine kinase that can promote glycolytic metabolism by activating hexokinase (HK) and phosphofructokinase (PFK), and increase glycolytic activity through Akt-mediated membrane translocation of GLUT. In addition, AKT activates a number of key downstream effectors that play an important role in cellular metabolic reprogramming, including mammalian target of rapamycin complex 1 (mTORC1), glycogen synthase kinase 3 (GSK3) and members of the forkhead box O (FOXO) family of transcription factors. HIF-1, which is stabilized in hypoxic circumstances and destroyed oxygen-dependently, can be stimulated by the multiprotein complex mTORC1. HIF-1 then stimulates the expression of GLUT1 and nearly all of the glycolysis-related enzymes. HIF1α also activates the expression of lactate dehydrogenase 1 (LDH1) and pyruvate dehydrogenase kinase 1 (PDK1), which together function to direct pyruvate to lactate instead of oxidizing it to acetyl coenzyme and entering the mitochondrial TCA cycle. Through the activation of the sterol regulatory element-binding protein 1 (SREBP1), hyperactive PI3K-AKT-mTORC1 signaling increases the expression of rate-limiting enzymes in the PPP (Glucose-6-phosphate dehydrogenase [G6PD]). Transketolase enzyme in the PPP is directly phosphorylated by aberrant AKT activation in cancer cells. Transcription factor c-Myc is one of the most hyperactivated genes in cancer cells that could be upregulated by growth factors (GF), mTOR, and transcription factor aryl hydrocarbon receptor (AHR). Myc increases glucose uptake and metabolism by inducing glycolytic enzymes (e.g., GLUT, HK2, LDH, and PFK) and rate-limiting enzymes of PPP (e.g., G6PD and TKT).

Lipid metabolism in tumors

A wide range of lipids are produced by the various classes of molecules known as fatty acids (FASs), which are made up of hydrocarbon chains with varying lengths and degrees of desaturation. In addition to cholesterol, FASs make up the hydrophobic tails of phospholipids and glycolipids, which together make up the majority of biological membranes. Additionally, the synthesis of steroid hormones and FAst-soluble vitamins uses cholesterol as a raw material. Second messengers, which are key signaling molecules for numerous biological actions, are also produced by membrane lipids. FASs are esterified with glycerol groups to form triacylglycerol (TAG), which are involved in the storage and release of energy. Thus the regulation of lipid metabolism is essential for the maintenance of cellular homeostasis.

In recent years, with the continuous deepening of research, there is growing awareness of the significance of FAS in the development, survival, and metastasis of cancer. Numerous studies have shown that the upregulation of FAS synthesis is a typical metabolic alteration in cancer, which is achieved by the upregulation of transporter proteins and various lipogenic enzymes. Mammals produce only certain FASs and other FASs notably polyunsaturated FASs, are acquired from the diet by FAS protein transporters (e.g., CD36, SLC27, and plasma membrane FAS-binding proteins [FASBPs]). Hydrogen sulfide, palmitic acid, and high-FAst diet-induced CD36 expression, which enhanced FAS uptake and cancer development. ATP-citrate lyase (ACLY), catalyzes the conversion of citrate and CoA to oxaloacetate and acetyl-CoA, is transcriptionally upregulated by sterol regulatory element binding protein 1 (SREBP1). AKT enhances lipid synthesis via direct phosphorylation of ACLY or induction of mTORC1, which enhances the translation and cleavage processing of SREBP FAsmily. AKT-mediated GSK3 inhibition prevents degradation of SREBP1. The mTOR signaling or SREBP can increase FAstty acid desaturase 2 (FASDS2) and stearoyl-CoA desaturase (SCD) expression, which are necessary for the generation of polyunsaturated FAstty acid (PUFAS) and monounsaturated FAstty acid (MUFAS), that prevent lipotoxicity and indirectly promotes cell proliferation, migration, and invasion. Acetyl-CoA synthetase2 (ACSS2), which is upregulated by SREBP and essential for converting acetate to acetyl-CoA, is expressed in a significant part of human malignancies. Acetyl-CoA carboxylase alpha (ACACA), the rate-limiting enzyme for FAS synthesis, catalyzes the carboxylation of acetyl-CoA to malonyl-CoA. It is regulated by SREBP at the transcriptional level and by AMP-activated protein kinase (AMPK) phosphorylation at the protein level. When aerobic glycolysis prevents glucose from entering the TCA cycle influx, glutamine acts as the primary carbon source in the TCA cycle and FAstty acid synthesis. Myc indirectly increases FAstty acid synthesis by enhancing glutamate-mediated TCA cycle influx through activation of GLUD.

The mevalonate pathway, which produces cholesterol, a crucial molecule for cell membrane function as well as other significant chemicals that might support tumor development and progression, is another typical lipogenic pathway that is activated in cancer. Mammalian 3-hydroxy-3-methylglutaryl (HMG)–CoA reductase (HMGCR), a rate-limiting enzyme for cholesterol biosynthesis in the mevalonate pathway, converts HMG-CoA to mevalonate. Squalene monooxygenase (SM), the second rate-limiting, transforms squalene to 2,3(S)-oxidosqualene. Both can be activated by SREBP2 to promote cholesterol synthesis.

Amino acid metabolism in tumors

Amino acids can be divided into essential amino acids (EAAs), which cannot be synthesized or the rate of synthesis is FAsr from adequate to the needs of the body and must be supplied by food proteins, and non-EAAs (NEAAs), which can be synthesized from simple precursors and do not need to be obtained from food. Because EAAs and NEAAs can be involved in metabolism in a variety of ways, including as a source of energy, substrate, and product of biomolecular transformations, and intermediate in redox reactions, the role of amino acid metabolism in cancer research has attracted increasingly attention of researchers.

The TCA cycle is powered by glutamine in cancer cells, which acts as an anaplerosis metabolite and sustain mitochondrial ATP synthesis through a series of biochemical reactions known as glutaminolysis. The α-KG, produced during glutaminolysis, can be converted to acetyl coenzyme for new FAstty acid synthesis. Glutamine acts as carbon and nitrogen donor for the synthesis of reduced glutathione (GSH), the latter contributes to the maintenance of redox balance. It also functions as nitrogen donor converting to other amino acids by transaminases or amidotransferases. In addition, glutamine provides an amide (γ-nitrogen) group that enables the de novo synthesis of nucleotides, which acts as an important role in cancer cell proliferation. Serine and glycine play a significant role in the development of cancer as one-carbon sources in DNA methylation and nucleotide synthesis. The branched-chain amino acids (BCAA; valine, leucine, and isoleucine) are EAAs for mammals. BCAAs cannot be synthesized by human cells, but are obtained via dietary intake and scavenged protein recycling. Highly reversible enzymes catabolize intracellular BCAAs to produce nitrogen and carbon groups for biomass synthesis, energy production, nutrient signaling, and epigenetic regulation.

Glutamine is imported into the cytoplasm via the glutamine transporter. c-Myc, mTORC1, and YAP/TAZ upregulate the expression of amino acid transporters (e.g., SLC1A5 and SLC7A5) and promote glutamine uptake. Amino acid transporters (e.g., SLC7A5 and SLC38A9) expression activates mTORC1 in turn. c-Myc activation inhibits miR-23a/b and promotes Glutaminase (GLS1/2) translation, the latter catalyzes glutamine to glutamate and produce ammonia (NH3). Under glutamine restriction, c-MYC can also boost demethylation of the glutamine synthetase (GS) promoter and upregulate expression of GS. KRAS and YAP/TAZ induce the aspartate synthase (GOT1) expression, which provides NADPH for the proper maintenance of redox homeostasis and produces NEAAs and TCA cycle intermediates for cell growth and produces NEAAs and TCA cycle intermediates. mTORC1 and mTORC2 regulate glutaminolysis by inducing the expression of glutamate dehydrogenase and GS, respectively.

Specific metabolic signatures for different cancers

In summary, cancer cells undergo a multiFAsceted metabolic remodeling to meet their growth requirements. However, not all cancer cells undergo the same metabolic remodeling. They may undergo a tissue-specific metabolic remodeling. Lung cancer is mainly divided into small cell lung cancer (SCLC) and non-SCLC (NSCLC). It is a multiFAsctorial disease with a high incidence and FAstality rate globally. Among these FAsctors, smoking is one of the main risk FAsctors, accounting for 25% of cases. In addition to the cellular metabolites, it is also necessary to examine metabolites associated with smoking. According to a study, cotinine significantly increases the risk of lung cancer. One of the major metabolic products of nicotine is cotinine, which accounts for 70%–80% of the metabolites produced in smokers. The liver enzymes CYP2A6, UDP-glucuronosyltransfease (UGT), and flavin-containing monooxygenase (FMO) are principally responsible for cotinine metabolism. Phosphoenolpyruvate carboxylase (PEPCK or PCK) is the first rate-limiting enzyme for gluconeogenesis, catalyzing the production of oxaloacetate to phosphoenolpyruvate to maintain glucose homeostasis. Mammalian cells encode two PCK genes, PCK1 and PCK2, PCK1 is found in the cytoplasm and PCK2 is found in the mitochondria, respectively. Both PCK genes are overexpressed in colon cancer, while the PCK2 gene is overexpressed in bladder, breast, kidney, and non-small cell lung cancer. Surprisingly, the expression of PCK1 and PCK2 were both significantly downregulated in hepatocellular carcinoma (HCC), indicating that gluconeogenic enzymes may play different roles in nongluconeogenic and gluconeogenic tissues. Endogenous estrogens and their metabolism, linked to breast carcinogenesis, play significant heterogeneity in breast cancer. Estrogen receptor α (ERα), once activated by estrogens, is responsible for the metabolic reprogramming, including aerobic glycolysis, nucleotide and amino acid synthesis, and choline (Cho) metabolism. Estrogen promotes consuming more glucose and producing more lactate, leading to a higher lactate/glucose ratio in cell. When ERα is stimulated by estrogen, choline phosphotransferase CHPT1 and phosphatidylcholine are both upregulated. High expression of CD36, a FAstty acid receptor and transporter, enhances the progression of solid malignancies, such as breast, ovarian, gastric, and glioblastoma. However, it has been shown that in colorectal cancer, activation of CD36 inhibits the proliferation of CRC cells and induces apoptosis. Helicobacter pylori (H. pylori) has long been recognized as class I carcinogen for gastric cancer. It was discovered that H. pylori infection caused the typical changes in amino acid levels in AGS cells. The levels of certain amino acids tended to be lower in the H. pylori-infected AGS cells than in the uninfected AGS cells at 2 h postinfection. However, the opposite tendency was observed at 6 and 12 h after infection. A study discovered 145 glycerophospholipids among 712 metabolites in all tumor/node/metastasis (TNM) esophageal squamous cell cancer (ESCC) and adjacent normal tissues, suggesting that glycerophospholipid metabolism was tightly linked to ESCC.

Metabolic reprogramming is a hallmark of cancer and a complex process. Both intrinsic and extrinsic FAsctors can lead to the reprogramming of cancer cells. However, the above article does not provide a detailed review of metabolic reprogramming in cancer cells, and readers are encouraged to refer to other literature for more information. Learning more about the molecular processes underlying metabolism can help to generate fresh suggestions for the early detection, prevention, and treatment of cancer.

APPLICATION OF NONTARGETED METABOLOMICS IN ONCOLOGY RESEARCH

As mentioned above, cancer is mainly caused by mutations in oncogenes. Changes of DNA lead to alter RNA transcription, protein expression, and protein function, which can catalyze the production or act on metabolites and ultimately cause corresponding changes in metabolites. In other words, genomics and proteomics tell you what could happen, while metabolomics tells you what has happened. Different physiopathological states exist at different metabolome levels, and in the same way, specific physiological or disease states can be determined by detecting metabolites. In some cases, it may be the most sensitive method for identifying pathological variants, since even small changes in protein expression or structure can lead to significant changes in protein activity and metabolite levels. In addition, a wide variety of samples are available for metabolomics, including tissues and biofluids. Compare to other omics, metabolomics could more accurately reflect the status of organism under drug or environmental influences. Therefore the use of metabolomics for cancer diagnostic biomarker screening is receiving increasing attention.

Screening for diagnosis and prognosis biomarkers

In recent years, many studies have identified metabolites that are used to predict cancer risk and prognosis. Acylcarnitine (C2) and phosphotidylcholines (PC ae C36:3) were strongly associated with the risk of breast cancer with the odd ratio (OR) 1.15 (95% confidence interval [95% CI]: 1.06–1.24) and 0.88 (95% CI: 0.81–0.95), respectively, and these were unaffected by breast cancer subtype, age at diagnosis, FAssting status, menopausal status, or adiposity. In a study, increased serum levels of N1, N12-diacetylspermine (DAS) aided to early detection of NACLC. For serum samples collected 6 to 12 months prediagnosis, the AUC was 0.610 (95% CI, 0.510–0.710), while for samples collected 0–6 months prediagnosis, it was 0.710 (95% CI: 0.613–0.808). Additionally, DAS levels were noticeably higher in samples taken 0 to 6 months prediagnosis than 6–12 months (p = 0.011), suggesting serum DAS levels not only can provide early diagnosis of NSCLC, but also predict tumor progression. A prospective analysis proved that pseudouridine (OR: 2.56, 95% CI: 1.48–4.45, p = 0.001, adjusted-p = 0.15) associated with ovarian cancer risk. Metabolomic analysis of gastric cancer (GC) cases and controls found that higher levels of 1-methylnicotinamide and N-acetyl-d-glucosamine-6-phosphate could be employed as a promising biomarker for GC tumor. Their corresponding AUCs were 0.957 (95% CI: 0.917–0.997) and 0.951 (95% CI: 0.901–1.000), respectively. The AUC for the combination of the two metabolites was 0.976 (95% CI: 0.940–1.000). In pancreatic cancer, researchers compared three metabolites (aspartate/alanine, androstenediol monosulFAste, and glycine) with a CA19-9-based reference model for ROC analysis, the AUC increased from 0.573 in the reference model to 0.721 (p = 4.8 × 10⁻⁴ by Delong's test), suggesting that these three biochemicals have potential value in distinguishing prediagnostic pancreatic cancer cases from controls. Monocarboxylate transporters 4 (MCT4) associated with lymph node metastasis (OR: 1.87, 95% CI: 1.10–3.17, p = 0.02) and distant metastasis (OR: 2.18, 95% CI:1.65–2.86, p < 0.001). The biomarker panel consisting of 3-chlorotyrosine, 12:0-carnitine, glutamate and phosphatidylcholine was able to distinguish between benign disease and lung adenocarcinoma (LUAD) group (adenocarcinoma in situ [AIS]/minimally invasive adenocarcinoma [MIA]/invasive adenocarcinoma [IAC]) with an AUC of 0.894. The combination of two metabolites (i.e., cystine and valine) showed a high AUC value of 0.865 in the classification of benign disease and AIS/MIA. Two groups of metabolites (i.e., asparagine and cystine) reached an AUC of 0.931 in distinguishing AIS from benign disease. Phosphatidylserine Synthase 1 (PTDSS1) and Lysophosphatidylcholine Acyltransferase 1 (LPCAT1) showed a good prediction of ESCC with an AUC of 0.980 and 0.914, respectively. These findings suggest that metabolomics can be used as a biomarker for the early detection and diagnosis of cancer. In addition, research on cancer metabolism not only provides a basis for diagnosis but also reveals the prognosis. According to an analysis of clinical outcome data in HER2+ breast cancer, high dihydroceramide desaturase 1 (DES1) expression was associated with lower relapse-free survival, overall survival, and distant metastasis-free survival compared to low DES1 tumors. The hazard rate (HR) was 2.36, 2.88, and 5.13, respectively. Overall survival (OS) was negatively associated with high GLUT1 expression (OS) in liver hepatocellular carcinoma (HR: 1.9, p = 7e−04). In cervical and endocervical cancer, patients with high expression of monocarboxylate transporter 4 (MCT4) showed a poor prognosis (HR: 2.0, p = 0.01). The increased G6PD associated with poor OS in kidney renal papillary cell carcinoma (HR: 2.4, p = 1e−03). In kidney renal clear cell carcinoma, low expression of FAstty acid transport proteins 2 was associated with poor prognosis (HR: 2.8, p = 4e−10). A meta-analysis was conducted on approximately 21,000 unique metabolites in the blood, urine, and tumor tissue of 1900 individuals across five cohorts, 300 tumor types. Vote counting statistics (VCS) is used to represent specific metabolites that continue to rise or FAsll in cancer. Lactate levels were elevated in all cancer types studied (VCS = 26/26, p = 1.5 × 10⁻⁶). Glutamate (VCS = 16/18, p = 3.6 × 10⁻³) and glutamine (VCS = 7/13, p = 1.1 × 10⁻¹), a glutamate precursor, were increased in tumor tissue. Glutamate was the most increased metabolite in the blood (VCS = 11/15, p = 5.5 × 10⁻²), and the ketone body 3-hydroxybutyric acid was increased (VCS = 9/15, p = 1.3 × 10⁻¹). Glutamine (VCS = −18/26, p = 8.0 × 10⁻³) tryptophan (VCS = −22/26, p = 3.2 × 10⁻⁴) and histidine (VCS = −14/18, p = 1.3 × 10⁻²) were decreased in the blood. We have summarized the biomarkers that are important for tumor diagnosis and prognosis in the last few years of research (Table ).

Table 1 Summary of recent metabolomics studies for breast cancer, lung cancer, hepatocellular carcinoma, colorectal cancer, gastric cancer, and esophageal cancer.

Cancer type	Specimen type	Metabolite	Value	References
Breast cancer	Tissue	Spermidine; kynurenine	Prognosis	[]
Breast cancer	Tissue	Ceramides; FAstty acids	Diagnosis	[]
Breast cancer	Plasma	Ethyl (R)−3-hydroxyhexanoat; caprylic acid; hypoxanthine	Diagnosis	[]
Breast cancer	Saliva	Arginase	Diagnosis	[]
Breast cancer	Serum	1-oleoylglycerophosphocholine	Diagnosis	[]
Breast cancer	Serum	Aspartate	Diagnosis	[]
Breast cancer	Serum	l-leucine; l-proline; l-threonine; l-tyrosine; pyroglutamic acids	Diagnosis	[]
Breast cancer	Exosomal	Succinate; lactate	Prognosis	[]
Lung cancer	Serum	7-alpha-hydroxy−3-oxo-4-cholestenoate	Diagnosis	[]
Lung cancer	Plasma	Palmitic acid; heptadecanoic acid; 4-Oxoproline; tridecanoic acid; Ornithine	Diagnosis	[]
Lung cancer	Serum	Glycerophospholipids; imidazopyrimidines	Diagnosis	[]
Lung cancer	Tissue	Adenine; histamine; inosine	Prognosis	[]
Lung cancer	Tissue	Guanine; choline; creatine	Prognosis	[]
Lung cancer	Serum	Xanthine; hypoxanthine; pyruvate; lactate	Diagnosis	[]
Lung cancer	Tissue	Isoleucine; creatinine; serine	Prognosis	[]
Lung cancer	Serum	Phenylalanylphenylalanine	Diagnosis	[]
Hepatocellular carcinoma	Serum	Phenylalanyl-tryptophan; glycocholate	Prognosis	[]
Hepatocellular carcinoma	Cell	Pro-CoA; propionyl-l-carnitine; 2-methylcitric acid	Diagnosis	[]
Hepatocellular carcinoma	Plasma	PC(16:0/16:0); PC(18:2/18:2); SM(d18:1/18:1)	Diagnosis and prognosis	[]
Hepatocellular carcinoma	Plasma	MG(18:2/0:0/0:0)	Prognosis	[]
Hepatocellular carcinoma	Tissue	2-hydroxystearate	Prognosis	[]
Hepatocellular carcinoma	Serum	Taurodeoxy cholic acid; 1,2-diacyl-3-β-d-galactosyl-sn-glycerol	Diagnosis	[]
Hepatocellular carcinoma	Tissues	Nicotinamide riboside; 4-hydroxyglutamate, deoxy-carnitine; S-adenosylmethionine, inosine 5′-monophosphate; guanosine 5′-onophosphate; homoserine	Prognosis	[]
Hepatocellular carcinoma	Tissue	Glucose; phosphoethanolamine; triacylglyceride	Prognosis	[]
Hepatocellular carcinoma	Serum	dl-3-phenyllactic acid; glycocholic acid; 1-methylnicotinamide; l-tryptophan	Diagnosis	[]
Colorectal cancer	Serum	Hexadecanedioic acid; 4-dodecylbenzenesulfonic acid; 2-pyrocatechuic acid; formylanthranilic acid	Diagnosis	[]
Colorectal cancer	Serum	Culinariside; 2-octenoylcarnitine; N, O-Bis- (trimethylsilyl)phenylalanine	Diagnosis	[]
Colorectal cancer	Plasma	C36:3 phosphatidylcholine; C54:8 triglyceride; phenylacetylglutamine	Diagnosis	[]
Colorectal cancer	Tissue	Glycerophospholipids; sphingomyelin; triacylglycerol; lipogenic enzymes	Prognosis	[]
Colorectal cancer	fecal	N-acetylmannosamine; 2,5-dihydroxybenzaldehyde	Prognosis	[]
Colorectal cancer	fecal	Succinic acid; aminoisobutyric acid; butyric acid; isoleucine; and leucine	Diagnosis	[]
Gastric cancer	Plasma	α-linolenic acid; linoleic acid; palmitic acid; arachidonic acid; sn-1 lysophosphatidylcholine (LysoPC)(18:3); sn-2 LysoPC(20:3)	Prognosis	[]
Gastric cancer	Tissue	1-methylnicotinamide; N-acetyl-d-glucosamine-6-phosphate	Diagnosis	[]
Gastric cancer	Plasma	Trimethylamine N-oxide	Diagnosis	[]
Gastric cancer	Serum	Hexadecasphinganine; linoleamide; N-hydroxy arachidonoyl amine	Diagnosis	[]
Gastric cancer	Tissue	Glycerophospholipid	Prognosis	[]
Gastric cancer	Tissue	5′-Methylthioadenosine; linoleic acid	Prognosis	[]
Gastric cancer	Serum	Pyroglutamic acid	Diagnosis	[]
Esophageal cancer	Tissue	Glycerophospholipid	Diagnosis	[]
Esophageal cancer	Serum	Glycine; fructose; ornithine; threonine	Prognosis	[]
Esophageal cancer	Serum	MG(20:4)isomer; 9,12-octadecadienoic acid; l-isoleucine	Prognosis	[]
Esophageal cancer	Serum	Acetate; pyruvate	Diagnosis	[]
Esophageal cancer	Tissue	Ganglioside GM2; oleoyl glycine	Diagnosis	[]
Esophageal cancer	Tissue	Adenylsuccinic acid, UDP-GalNAc, maleylacetoacetic acid, hydroxyphenylacetylglycine, galactose, kynurenine	Diagnosis	[]
Esophageal cancer	Tissue	2-hydroxymyristoylcarnitine, hydroxyhexadecanoylcarnitine, 2,3-Dinor-TXB1	Prognosis	[]

Pan-cancer metabolomics analysis

Mullen et al. shed light on the aberrant regulation of nucleotide metabolism significantly impacts diverse cancer cell activities, including growth, immune evasion, metastasis, and resistance to cancer therapy. This highlights the role of aberrant nucleotide synthesis as a universal metabolic dependency in pan-cancer. Choi et al. conducted a comprehensive investigation of pan-cancer metabolic landscapes to identify shared patterns and associations with cancer progression. Their study revealed commonalities among different types of cancer regarding increased metabolic activity in carbohydrates and nucleotides. These metabolic alterations were found to be associated with poor prognosis and tumor mutational load (TMB), with ribonucleotide synthesis emerging as the most prominent feature followed by pyrimidine, purine, and carbohydrate metabolism. These findings support the idea of a link between metabolic landscape remodeling and clinical progression of cancer. In a comprehensive study, Daemen et al. found a high degree of codependence between glutaminase (GLS1) and γ-glutamylcysteine synthetase (GCS) inhibitors in 407 tumor lines across various indications. Through validation of their findings, they derived a pan-cancer metabolic signature that predicts GLS1/GCS codependence. Rohatgi et al. analyzed the metabolic profile of cancer and noncancer cells (stroma) in bulk tumors of 20 solid tumor types. The results showed that oxidative phosphorylation was the most upregulated metabolic process in cancer cells and that the rate-limiting enzymes of tryptophan metabolism, IDO1 and TDO2, were highly overexpressed in stroma. Other studies demonstrated that tryptophan catabolism is upregulated in stromal cells and is becoming a target for cancer immunotherapy. It is shown that by codiscovering the metabolic characteristics of cancer cells and stromal cells in tumor types, it can provide a basis for studying the reciprocal metabolic roles between malignant and nonmalignant cells in the tumor microenvironment (TME).

It is well known that the tumor microenvironment is the internal environment in which tumor cells arise and live, and it is heterogeneous and consists of multiple cell types, playing a key role in the pathogenesis of cancer. The above studies found that pan-cancer research can not only discover the metabolic characteristics of cancer cells but also study the metabolic characteristics of the stroma, which provides a new idea to understand the complex interactions between the systemic mediators of disease progression for the rational development of effective anticancer therapies.

Single-cell gene analysis on the metabolic pathways

Single-cell gene expression analysis of inferred malignant and nonmalignant cells can also be used to study the heterogeneity of metabolic pathways within tumors. Single-cell RNA-seq (scRNA-seq) data have been widely used to characterize cell type-specific transcriptional states and their potential phenotypic transitions in complex tissues. More and more scRNA-seq-related software and strategies have been developed to improve the accuracy and dimensionality of bioinformatics analyses, and to achieve a tight connection between transcriptome and metabolome profiles. Wu et al. developed scMetabolism, a computational pipeline for quantifying single-cell metabolism, and identified highly metabolically activated MRC1 + CCL18 + M2-like macrophages at metastatic sites of colorectal cancer. Highly effective neoadjuvant chemotherapy can slow down this metabolic activation, thus opening up the possibility of targeting metabolic pathways in metastasis. Damiani et al. proposed a computational framework for single-cell flux balance analysis (scFBA), which can transform the single-cell transcriptome into a single-cell fluxome to characterize metabolism. Wagner et al. present compass, an algorithm for characterizing cellular metabolic states based on single-cell RNA sequencing and flux balance analysis. It is used to analyze immune cell metabolism (immunometabolism). The researchers applied Compass to correlate metabolic status with helper T-cell 17 (Th17) functional variability (pathogenic potential) and recovered metabolic transitions between glycolysis and FAstty acid oxidation, which resembled known Th17/regulatory T cell (Treg) differences and were validated by metabolic profiling Compass also predicted that Th17 pathogenicity is associated with arginine and downstream polyamine metabolism. Alghamdi et al. developed scFEA, a computational approach that utilizes a probabilistic model of flux equilibrium constraints based on scRNA-seq data to infer cellular metabolic fluxomes and changes in metabolite abundance at the single-cell level. The estimated cellular fluxomes can be used for downstream analyses such as identifying metabolic stresses, assessing the sensitivity of individual enzymes to metabolic networks, and inferring cell–tissue and cell–cell metabolic exchanges.

In the realm of pan-cancer metabolomics research, research relies heavily on transcriptomic data to infer metabolic activity and phenotypes in cancer and stromal cells. While these data offer valuable insights into cellular metabolic activity, the confirmation, and validation of metabolic pathways in these cells necessitate the inclusion of metabolite-level data. Therefore, to further substantiate the findings, additional information on metabolite levels becomes imperative.

Resources and tools for pan-cancer metabolomics analysis

The current demand for pan-cancer analysis requires the integration of a substantial amount of metabolomics data, which unfortunately, has been gathered and processed across various platforms and implemented with different pipelines. This heterogeneity in data collection and analysis poses challenges in terms of reproducibility and comparability, thus hindering comprehensive investigations. Publics repositories providing valuable resources for reusing metabolomics data beyond the text in manuscripts. MetaboLights is a public metabolome database hosted by EMBL-EBI (European Molecular Biology Laboratory—European Bioinformatics Institute). It stores raw metabolomic data obtained from different species and different technologies and is recommended as a repository for publication. Currently, MetaboLights includes 1197 studies and provides valuable resources for reproduce and integration research. Metabolomics Workbench is another public repository for cross-species and cross-platform metabolomic experimental data. It provides comprehensive tools for integrate, analyze, track, deposit, and disseminate the metabolomic data. Currently, it covers over 20 species and includes 1988 studies.

Except public repository, knowledge database is also essential for metabolomics analysis. HMDB (The Human Metabolome Database) is a reference database of human metabolites, providing comprehensive information including biological, physiological, and chemical properties. Currently it includes 217,920 compounds and provides more accurate predicted NMR spectra and MS spectra. METLIN is another reference database that includes reference information on metabolites. METLIN provides a large amount of mass spectrometry information and useful for metabolite identification. KEGG (Kyoto Encyclopedia of Genes and Genomes) is a comprehensive database mainly focus on pathway profiling. With the pathway map in EKGG, research could annotate and understanding the understanding high-level functions of metabolic changes in the biological system. LIPID MAPS is a database specified for lipidomic and provides more precisely classification of lipids. TCGA (The Cancer Genome Atlas) is a comprehensive “atlas” of cancer genomic profiles. ICGC (International Cancer Genome Consortium) collected tumor data from 50 different cancer types (or subtypes), including abnormal gene expression, somatic mutations, epigenetic modifications, clinical data, and so forth. GTEx (Genotype-Tissue Expression), establish a resource database and associated tissue bank. All these databases provide web-based tools for browse and analysis (Table ).

Table 2 Resources and tools for metabolomics analysis.

Type	Name	Website	References
Public repositories
	EBI MetaboLights		[]
	Metabolomics Workbench		[]
Knowledge databases
	HMDB		[]
	METLIN		[]
	KEGG		[]
	LIPID MAPS		[]
	TCGA		[]
	ICGC		[]
	GTEx		[]
Analysis tools
	xcms		[]
	MS-DIAL		[]
	MetDNA		[]
	MetaboAnalyst		[]
	TidyMass		[]

Bioinformatic analysis of metabolomics is tricky and plenty tools have been developed for certain step. Here we summarize several tools that designed for generate workflows for the analysis of metabolomics data. Unlike tools for certain usage, workflow tools can provide interconnected tools and make it ease-of-use. XCMS includes online tools and R packages and incorporate the entire Liquid chromatography–mass spectrometry data analysis workflow. MD-DIAL is a noncommercial software for data-independent MS/MS (tandem mass spectrometry) data deconvolution. MetaboAnalyst is webtools for LC-HRMS (high-resolution mass spectrometry) data analysis. In the current version, it optimizes parameters for LC-HRMS and provides more tools for functional analysis and data integration. MetDNA is web-based tool that provides a unique metabolites annotation method named KGMN (knowledge-guided multi-layer network), which uses known biological knowledge to improve the annotation accuracy. Tidymass is a R package that provides an object-oriented reproducible analysis framework, which increased the reproducibility of analysis.

Applications of machine learning (ML) in pan-cancer metabolomics

Mass spectrometry metabolomics, as one of the high-throughput biological experiments, also generates a large amount of structurally complex data. These data are characterized by high noise, high dimensionality, complex interactions, and redundancy. Before using mass spectrometry data for analysis, a series of processing processes such as data preprocessing, peak extraction, peak detection, peak alignment, and metabolite identification are required. In addition, the analysis of metabolomics data also FAsces the problems of batch effects in the measurement process and the integration with other histological data. Therefore, a high level of statistical and computational methods analysis is needed to help reveal the underlying trends and characteristics.

In recent years, ML has been applied in various fields of biomedicine with its powerful ability to solve nonlinear relationships and handle large heterogeneous data sets. demonstrating the great promise of ML in disease prediction and diagnosis. The general steps of ML are as follows: (1) Data collection: Collect data sets and preprocess them according to the task. (2) Feature engineering: Extract relevant features from the data sets and process and transform them. (3) Data partition: Split the data sets into training, validation, and test sets. (4) Model selection: Choose a suitable model based on the task, select algorithms, adjust hyperparameters, and so forth. (5) Model training: Train the model using the training set and optimize it through the validation set. (6) Model evaluation: Evaluate the performance of the model using the test set. (7) Model deployment: Apply the trained model to practical scenarios. (8) Model optimization: Optimize the model based on the actual effect and repeat the above steps (Figure ).

[IMAGE OMITTED. SEE PDF]

As is well known, metabolic data need to be standardized before statistical analysis to eliminate differences between different batches and instruments, including peak extraction, peak detection, peak alignment, missing value filling, metabolite annotation, and other steps. DL methods have been proposed for critical data preprocessing steps. Ji et al. presented a novel method termed deep-learning-based sequential window acquisition of all theoretical spectra (DeepSWATH), which utilizes a spectral-centric approach for the identification of nontargeted metabolites from SWATH-MS data sets. The DeepSWATH methodology is based on the establishment of a precursor-fragmentation correlation (PFC) model, which utilizes convolutional neural networks for the extraction of fragments of precursor ions. The MS/MS spectra obtained through the DeepSWATH approach are superior in accuracy compared to those obtained with MS-DIAL. These results also display metabolite identification consistency in comparison to DDA MS/MS spectra. Furthermore, DeepSWATH can handle coelution condition. Gloaguen et al. proposed NeatMS based on CNN ML approach, which reduces the number and proportion of FAslse peaks. When it comes to model training, there are typically three main approaches: reinforcement learning, unsupervised learning, and supervised learning. The main machine algorithms in metabolomics are supervised learning and unsupervised learning. Supervised learning methods mainly include linear regression, logistic regression, Linear Discriminant Analysis (LDA), k-Nearest Neighbers (k-NN), Support Vector Machines (SVM), decision tree and Random Forest (RF), partial least-squares discrimination analysis (PLS-DA), orthogonal partial least-squares discrimination analysis (OPLS-DA), neural networks (NN), and other methods. Examples of unsupervised algorithms include principal component analysis (PCA), kernel PCA, t-distributed stochastic neighbor embedding (t-SNE), uniform manifold approximation and projection (UMAP), k-means Clustering, apriori, hierarchical clustering, and so on. In the study of early detection of lung adenocarcinoma and classification of lung nodules, Wang et al. utilized the t-SNE method to perform dimensionality reduction analysis on metabolic data. Additionally, they conducted diagnostic analyses on lung adenocarcinoma using several ML methods, including logistic regression, elastic net (EN) least absolute shrinkage and selection operator (LASSO), SVM, and RF. Furthermore, they constructed two NN models based on deep learning for lung adenocarcinoma diagnosis and lung nodule classification. Guo et al. utilized Lasso for feature selection and conducted 10-fold cross-validation in their research. They employed ML methods such as random forest analysis, support vector machines, multilayer perceptron, and CNN for model development. Among these models, random forest performed well and was identified as the best model for early diagnosis of kidney disease.

Currently, ML has been widely applied in metabolomics, with new algorithms and models continuously emerging. Its ability of end-to-end prediction might provide new tools for identification and analysis. Selecting the appropriate ML algorithm is crucial for the success of metabolomics research. This requires researchers to have a good understanding of the advantages of each ML method, the nature of their data, and the experimental objectives, and to select the most appropriate method that fits their needs to obtain reliable and interpretable results.

CONCLUSIONS AND PERSPECTIVE

Despite being a relatively nascent discipline. metabolomics has generated a multitude of biologically relevant findings and has unveiled bioactive metabolites across diverse domains of life. This underscores the immense potential and wide-ranging applicability of metabolomics. In recent years, there has been a growing number of reported pan-cancer analysis projects focused on identifying common molecular characteristics and functional genes associated with specific biological pathways across various tumor types. These studies have greatly contributed to a multidimensional, thorough, and comprehensive comprehension of human cancers. However, it is noteworthy that the number of studies exploring pan-cancer metabolism remains relatively limited in comparison.

As mentioned above, although a large amount of metabolomic data has been generated in different tumor types, these studies either have small data sets or focus on single tumors, preventing us from having a very comprehensive understanding of the disease as a whole. In FAsct, some common metabolic pathways and processes exist between multiple cancer types. For example, many cancer cells rely on glycolysis and FAS synthesis to meet their energy and structural unit requirements. Identifying these common metabolic pathways and processes can help guide the development of more broadly effective therapeutic approaches. Pan-cancer metabolism research seeks to identify common metabolic pathways and processes that are shared across different types of cancer. This approach can help to identify metabolic targets that are relevant to multiple cancer types, which can lead to the development of more broadly effective therapies.

Metabolomics has emerged relatively recently, bring new insight for cancer development and providing new biomarkers for diagnosis and prognosis. However, because the differences of samples, vendors, and pipelines in different studies, it's hard to integrate the results of current experiments especially in pan-cancer analysis. To solve this problem, one possible way is to create pan-cancer sample library and perform the metabolomics analysis follow a common pipeline. TCGA have down a great work on tumor genomics, epigenomics, and transcriptomes. Now the metabolomics research needs seem resources. Another problem is the metabolite identification in untargeted metabolomics. Since untargeted metabolome could systematic measure metabolites in a sample, however, the metabolites can't be accurate identified automatically, which were rely on manually curation. Besides the varies pipelines for data different vendor were also increase the limitation of certain identification tools.

In the future, we can work on expanding the sample set to cover different types of tumors, developing methods to integrate multiomics data, focusing on metabolic changes in the microenvironment, introducing quantitative methods, utilizing ML and artificial intelligence, and further exploring metabolite interactions (Figure ). By combining nontargeted metabolomics data with genomics, transcriptomics, and proteomics data, an integrated analysis framework can be established, leading to a more comprehensive understanding of the regulatory mechanisms of tumor metabolism and the identification of pan-cancer disease markers with greater accuracy and reliability. Additionally, the microenvironment of a tumor has a significant impact on the metabolism of tumor cells, and studying the application of nontargeted metabolomics in the tumor microenvironment and exploring the interactions of metabolites between tumor cells and the surrounding tissues or cells will help to gain a deeper understanding of the metabolic mechanisms of tumor development and progression. Moreover, improving and optimizing quantitative analysis methods in nontargeted metabolomics to more accurately measure metabolite concentrations and changes will help improve the specificity of identifying pan-cancer markers and provide more reliable results for clinical applications. Applying ML and artificial intelligence to the analysis and interpretation of nontargeted metabolomics data can accelerate the process of screening and identification of pan-cancer markers, improving the efficiency of diagnosis and treatment. Furthermore, investigating the interaction networks among metabolites to understand their relationships and regulatory mechanisms in tumor metabolic pathways can help to identify more specific pan-cancer markers and provide a basis for the development of new therapeutic strategies.

[IMAGE OMITTED. SEE PDF]

Although metabolomics is relatively less used than other omics, metabolomics data have the advantage of providing information on metabolite levels that can characterize disease stages. Pan-cancer studies have demonstrated great potential in the field of tumor metabolomics and have provided new perspectives and directions for our in-depth understanding of the development and progression of human cancers. As the research continues to deepen and develop, we expect to discover more common metabolic signatures and provide strategies with more practical applications for tumor diagnosis and treatment.

AUTHOR CONTRIBUTIONS

Taorui Wang: Conceptualization (equal); investigation (equal); resources (equal); writing—original draft (equal); writing—review and editing (equal). Yuanxu Gao: Conceptualization (equal); investigation (equal); resources (equal); writing—original draft (equal); writing—review and editing (equal). Both authors have read and approved the final manuscript.

ACKNOWLEDGMENTS

The figures created by Figdraw. This work was supported by the Science and Technology Development Fund, Macau SAR (0003-2021-AKP, SKL-LPS (MUST)−2021-2023).

CONFLICT OF INTEREST STATEMENT

Author Yuanxu Gao is an editorial staff of MedComm—Future Medicine. Author Yuanxu Gao was not involved in the journal's review of or decisions related to this manuscript. The other author declared no conflict of interest.

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

ETHICS STATEMENT

The authors have nothing to report.

References

Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68(6):394‐424.

Schiliro C, Firestein BL. Mechanisms of metabolic reprogramming in cancer cells supporting enhanced growth and proliferation. Cells. 2021;10(5):1056.

Warburg O, Wind F, Negelein E. The metabolism of tumors in the body. J Gen Physiol. 1927;8(6):519‐530.

Gandhi N, Das G. Metabolic reprogramming in breast cancer and its therapeutic implications. Cells. 2019;8(2):89.

Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144(5):646‐674.

Vander Heiden MG, DeBerardinis RJ. Understanding the intersections between metabolism and cancer biology. Cell. 2017;168(4):657‐669.

Pavlova NN, Thompson CB. The emerging hallmarks of cancer metabolism. Cell Metab. 2016;23(1):27‐47.

Lunt SY, Vander Heiden MG. Aerobic glycolysis: meeting the metabolic requirements of cell proliferation. Annu Rev Cell Dev Biol. 2011;27:441‐464.

Demetrius L, Tuszynski JA. Quantum metabolism explains the allometric scaling of metabolic rates. J R Soc Interface. 2010;7(44):507‐514.

Liberti MV, Locasale JW. The Warburg effect: how does it benefit cancer cells? Trends Biochem Sci. 2016;41(3):211‐218.

de Koning TJ, Snell K, Duran M, Berger R, Poll‐The BT, Surtees R. L‐serine in disease and development. Biochem J. 2003;371(Pt 3):653‐661.

Akella NM, Ciraku L, Reginato MJ. Fueling the fire: emerging role of the hexosamine biosynthetic pathway in cancer. BMC Biol. 2019;17(1):52.

Hui S, Ghergurovich JM, Morscher RJ, et al. Glucose feeds the TCA cycle via circulating lactate. Nature. 2017;551(7678):115‐118.

Cox AG, Tsomides A, Yimlamai D, et al. Yap regulates glucose utilization and sustains nucleotide synthesis to enable organ growth. EMBO J. 2018;37(22): [eLocator: e100294].

Ying H, Kimmelman AC, Lyssiotis CA, et al. Oncogenic Kras maintains pancreatic tumors through regulation of anabolic glucose metabolism. Cell. 2012;149(3):656‐670.

Osthus RC, Shim H, Kim S, et al. Deregulation of glucose transporter 1 and glycolytic gene expression by c‐Myc. J Biol Chem. 2000;275(29):21797‐21800.

Christofk HR, Vander Heiden MG, Harris MH, et al. The M2 splice isoform of pyruvate kinase is important for cancer metabolism and tumour growth. Nature. 2008;452(7184):230‐233.

Ediriweera MK, Tennekoon KH, Samarakoon SR. Role of the PI3K/AKT/mTOR signaling pathway in ovarian cancer: biological and therapeutic significance. Sem Cancer Biol. 2019;59:147‐160.

Alzahrani AS. PI3K/Akt/mTOR inhibitors in cancer: at the bench and bedside. Sem Cancer Biol. 2019;59:125‐132.

Elstrom RL, Bauer DE, Buzzai M, et al. Akt stimulates aerobic glycolysis in cancer cells. Cancer Res. 2004;64(11):3892‐3899.

Tanner LB, Goglia AG, Wei MH, et al. Four key steps control glycolytic flux in mammalian cells. Cell Syst. 2018;7(1):49‐62.e8.

Roberts DJ, Tan‐Sah VP, Smith JM, Miyamoto S. Akt phosphorylates HK‐II at Thr‐473 and increases mitochondrial HK‐II association to protect cardiomyocytes. J Biol Chem. 2013;288(33):23798‐23806.

Waldhart AN, Dykstra H, Peck AS, et al. Phosphorylation of TXNIP by AKT mediates acute influx of glucose in response to insulin. Cell Rep. 2017;19(10):2005‐2013.

Manning BD, Toker A. AKT/PKB signaling: navigating the network. Cell. 2017;169(3):381‐405.

Düvel K, Yecies JL, Menon S, et al. Activation of a metabolic gene regulatory network downstream of mTOR complex 1. Mol Cell. 2010;39(2):171‐183.

Cui X, Han Z, He S, et al. HIF1/2α mediates hypoxia‐induced LDHA expression in human pancreatic cancer cells. Oncotarget. 2017;8(15):24840‐24852.

Ghosh R, Samanta P, Sarkar R, et al. Targeting HIF‐1α by natural and synthetic compounds: a promising approach for anti‐cancer therapeutics development. Molecules. 2022;27(16):5192.

Dibble CC, Manning BD. Signal integration by mTORC1 coordinates nutrient input with biosynthetic output. Nature Cell Biol. 2013;15(6):555‐564.

Goul C, Peruzzo R, Zoncu R. The molecular basis of nutrient sensing and signalling by mTORC1 in metabolism regulation and disease. Nat Rev Mol Cell Biol. Published online August 23, 2023. [DOI: https://dx.doi.org/10.1038/s41580-023-00641-8]

Luo P, Yin P, Hua R, et al. A large‐scale, multicenter serum metabolite biomarker identification study for the early detection of hepatocellular carcinoma. Hepatology. 2018;67(2):662‐675.

Sun J, Ding J, Shen Q, et al. Decreased propionyl‐CoA metabolism facilitates metabolic reprogramming and promotes hepatocellular carcinoma. J Hepatol. 2023;78(3):627‐642.

Lin Z, Li H, He C, et al. Metabolomic biomarkers for the diagnosis and post‐transplant outcomes of AFP negative hepatocellular carcinoma. Front Oncol. 2023;13: [eLocator: 1072775].

Xia H, Chen J, Sekar K, Shi M, Xie T, Hui KM. Clinical and metabolomics analysis of hepatocellular carcinoma patients with diabetes mellitus. Metabolomics. 2019;15(12):156.

Pan HY, Wu QQ, Yin QQ, et al. LC/MS‐based global metabolomic identification of serum biomarkers differentiating hepatocellular carcinoma from chronic hepatitis B and liver cirrhosis. ACS Omega. 2021;6(2):1160‐1170.

Hou G, Ding D, Tian T, et al. Metabolomics‐based classification reveals subtypes of hepatocellular carcinoma. Mol Carcinog. 2022;61(11):989‐1001.

Buchard B, Teilhet C, Abeywickrama Samarakoon N, et al. Two metabolomics phenotypes of human hepatocellular carcinoma in non‐alcoholic fatty liver disease according to fibrosis severity. Metabolites. 2021;11(1):54.

Liu J, Geng W, Sun H, et al. Integrative metabolomic characterisation identifies altered portal vein serum metabolome contributing to human hepatocellular carcinoma. Gut. 2022;71(6):1203‐1213.

Zhang C, Zhou S, Chang H, et al. Metabolomic profiling identified serum metabolite biomarkers and related metabolic pathways of colorectal cancer. Dis Markers. 2021;2021:1‐9.

Chen F, Dai X, Zhou CC, et al. Integrated analysis of the faecal metagenome and serum metabolome reveals the role of gut microbiome‐associated metabolites in the detection of colorectal cancer and adenoma. Gut. 2022;71(7):1315‐1325.

Hang D, Zeleznik OA, Lu J, et al. Plasma metabolomic profiles for colorectal cancer precursors in women. Eur J Epidemiol. 2022;37(4):413‐422.

Ecker J, Benedetti E, Kindt ASD, et al. The colorectal cancer lipidome: identification of a robust tumor‐specific lipid species signature. Gastroenterology. 2021;161(3):910‐923.

Xie Z, Zhu R, Huang X, et al. Metabolomic analysis of gut metabolites in patients with colorectal cancer: association with disease development and outcome. Oncol Lett. 2023;26(2):358.

Yagin FH, Alkhateeb A, Colak C, Azzeh M, Yagin B, Rueda L. A fecal‐microbial‐extracellular‐vesicles‐based metabolomics machine learning framework and biomarker discovery for predicting colorectal cancer patients. Metabolites. 2023;13(5):589.

Huang S, Guo Y, Li ZW, et al. Identification and validation of plasma metabolomic signatures in precancerous gastric lesions that progress to cancer. JAMA Netw Open. 2021;4(6): [eLocator: e2114186].

Li B, Shu X, Jiang H, et al. Plasma metabolome identifies potential biomarkers of gastric precancerous lesions and gastric cancer risk. Metabolomics. 2023;19(8):73.

Yu L, Lai Q, Feng Q, Li Y, Feng J, Xu B. Serum metabolic profiling analysis of chronic gastritis and gastric cancer by untargeted metabolomics. Front Oncol. 2021;11: [eLocator: 636917].

Wang Y, Chen W, Li K, et al. Tissue‐based metabolomics reveals metabolic signatures and major metabolic pathways of gastric cancer with help of transcriptomic data from TCGA. Biosci Rep 2021;41(10): [eLocator: BSR20211476].

Cao K, Lyu Y, Chen J, et al. Prognostic implication of plasma metabolites in gastric cancer. Int J Mol Sci. 2023;24(16): [eLocator: 12774].

Li M, Zhang R, Xin M, et al. Discovery and validation of potential serum biomarkers with pro‐inflammatory and DNA damage activities in ulcerative colitis: a comprehensive untargeted metabolomic study. Metabolites. 2022;12(10):997.

Yang X, Wang P, Ye H, et al. Untargeted serum metabolomics reveals potential biomarkers and metabolic pathways associated with esophageal cancer. Front Oncol. 2022;12: [eLocator: 938234].

Wang PP, Song X, Zhao XK, et al. Serum metabolomic profiling reveals biomarkers for early detection and prognosis of esophageal squamous cell carcinoma. Front Oncol. 2022;12: [eLocator: 790933].

Ye W, Lin Y, Bezabeh T, et al. (1) H NMR‐based metabolomics of paired esophageal tumor tissues and serum samples identifies specific serum biomarkers for esophageal cancer. NBM. 2021;34(6): [eLocator: e4505].

Xu J, Cao W, Shao A, et al. Metabolomics of esophageal squamous cell carcinoma tissues: potential biomarkers for diagnosis and promising targets for therapy. BioMed Res Int. 2022;2022:1‐24.

Chen Z, Gao Y, Huang X, et al. Tissue‐based metabolomics reveals metabolic biomarkers and potential therapeutic targets for esophageal squamous cell carcinoma. J Pharm Biomed Anal. 2021;197: [eLocator: 113937].

Mullen NJ, Singh PK. Nucleotide metabolism: a pan‐cancer metabolic dependency. Nat Rev Cancer. 2023;23(5):275‐294.

Choi H, Na KJ. Pan‐cancer analysis of tumor metabolic landscape associated with genomic alterations. Mol Cancer. 2018;17(1):150.

Daemen A, Liu B, Song K, et al. Pan‐cancer metabolic signature predicts co‐dependency on glutaminase and de novo glutathione synthesis linked to a high‐mesenchymal cell state. Cell Metab. 2018;28(3):383‐399.

Rohatgi N, Ghoshdastider U, Baruah P, Kulshrestha T, Skanderup AJ. A pan‐cancer metabolic Atlas of the tumor microenvironment. Cell Rep. 2022;39(6): [eLocator: 110800].

de Visser KE, Joyce JA. The evolving tumor microenvironment: from cancer initiation to metastatic outgrowth. Cancer Cell. 2023;41(3):374‐403.

Wu Y, Yang S, Ma J, et al. Spatiotemporal immune landscape of colorectal cancer liver metastasis at single‐cell level. Cancer Discov. 2022;12(1):134‐153.

Damiani C, Maspero D, Di Filippo M, et al. Integration of single‐cell RNA‐seq data into population models to characterize cancer metabolism. PLoS Comput Biol. 2019;15(2): [eLocator: e1006733].

Wagner A, Wang C, Fessler J, et al. Metabolic modeling of single Th17 cells reveals regulators of autoimmunity. Cell. 2021;184(16):4168‐4185.

Alghamdi N, Chang W, Dang P, et al. A graph neural network model to estimate cell‐wise metabolic flux using single‐cell RNA‐seq data. Genome Res. 2021;31(10):1867‐1884.

Haug K, Cochrane K, Nainala VC, et al. MetaboLights: a resource evolving in response to the needs of its scientific community. Nucleic Acids Res. 2020;48(D1):440.

Sud M, Fahy E, Cotter D, et al. Metabolomics workbench: an international repository for metabolomics data and metadata, metabolite standards, protocols, tutorials and training, and analysis tools. Nucleic Acids Res. 2016;44(D1):D463‐D470.

Wishart DS, Guo A, Oler E, et al. HMDB 5.0: the human metabolome database for 2022. Nucleic Acids Res. 2022;50(D1):D622‐D631.

Xue J, Guijas C, Benton HP, Warth B, Siuzdak G. METLIN MS(2) molecular standards database: a broad chemical and biological resource. Nat Methods. 2020;17(10):953‐954.

Okuda S, Yamada T, Hamajima M, et al. KEGG Atlas mapping for global analysis of metabolic pathways. Nucleic Acids Res. 2008;36:W423‐W426.

Sud M, Fahy E, Cotter D, et al. LMSD: LIPID MAPS structure database. Nucleic Acids Res. 2007;35:D527‐D532.

Tomczak K, Czerwińska P, Wiznerowicz M. The Cancer Genome Atlas (TCGA): an immeasurable source of knowledge. Contemp Oncol (Poznan, Poland). 2015;19(1a):A68‐A77.

Zhang J, Baran J, Cros A, et al. International cancer genome consortium data portal—a one‐stop shop for cancer genomics data. Database. 2011;2011: [eLocator: bar026].

Lonsdale J, Thomas J, Salvatore M, et al. The genotype‐tissue expression (GTEx) project. Nat Genet. 2013;45(6):580‐585.

Forsberg EM, Huan T, Rinehart D, et al. Data processing, multi‐omic pathway mapping, and metabolite activity analysis using XCMS online. Nat Protoc. 2018;13(4):633‐651.

Tsugawa H, Cajka T, Kind T, et al. MS‐DIAL: data‐independent MS/MS deconvolution for comprehensive metabolome analysis. Nat Methods. 2015;12(6):523‐526.

Zhou Z, Luo M, Zhang H, Yin Y, Cai Y, Zhu ZJ. Metabolite annotation from knowns to unknowns through knowledge‐guided multi‐layer metabolic networking. Nat Commun. 2022;13(1):6656.

Pang Z, Zhou G, Ewald J, et al. Using MetaboAnalyst 5.0 for LC‐HRMS spectra processing, multi‐omics integration and covariate adjustment of global metabolomics data. Nat Protoc. 2022;17(8):1735‐1761.

Shen X, Yan H, Wang C, Gao P, Johnson CH, Snyder MP. TidyMass an object‐oriented reproducible analysis framework for LC‐MS data. Nat Commun. 2022;13(1):4365.

Galal A, Talal M, Moustafa A. Applications of machine learning in metabolomics: disease modeling and classification. Front Genet. 2022;13: [eLocator: 1017340].

Ji H, Lu H, Zhang Z. Deep Learning Enable Untargeted Metabolite Extraction from High Throughput Coverage Data‐Independent Acquisition. bioRxiv. 2020;1: [eLocator: 002683].

Gloaguen Y, Kirwan JA, Beule D. Deep learning‐assisted peak curation for large‐scale LC‐MS metabolomics. Anal Chem. 2022;94(12):4930‐4937.

Sidak D, Schwarzerová J, Weckwerth W, Waldherr S. Interpretable machine learning methods for predictions in systems biology from omics data. Front Mol Biosci. 2022;9: [eLocator: 926623].

Wang L, Zhang M, Pan X, et al. Integrative serum metabolic fingerprints based multi‐modal platforms for lung adenocarcinoma early detection and pulmonary nodule classification. Adv Sci (Weinheim, Baden‐Wurttemberg, Germany). 2022;9(34): [eLocator: 2203786].

Guo Y, Yu H, Chen D, Zhao YY. Machine learning distilled metabolite biomarkers for early‐stage renal injury. Metabolomics. 2019;16(1):4.

Sen P, Lamichhane S, Mathema VB, et al. Deep learning meets metabolomics: a methodological perspective. Brief. Bioinform. 2021;22(2):1531‐1542.

Alseekh S, Fernie AR. Metabolomics 20 years on: what have we learned and what hurdles remain? Plant J. 2018;94(6):933‐942.

Rinschen MM, Ivanisevic J, Giera M, Siuzdak G. Identification of bioactive metabolites using activity metabolomics. Nat Rev Mol Cell Biol. 2019;20(6):353‐367.

Cui K, Liu C, Li X, Zhang Q, Li Y. Comprehensive characterization of the rRNA metabolism‐related genes in human cancer. Oncogene. 2020;39(4):786‐800.

Wei C, Wang B, Peng D, et al. Pan‐cancer analysis shows that ALKBH5 is a potential prognostic and immunotherapeutic biomarker for multiple cancer types including gliomas. Front Immunol. 2022;13: [eLocator: 849592].

Chen G, Luo D, Zhong N, et al. GPC2 is a potential diagnostic, immunological, and prognostic biomarker in pan‐cancer. Front Immunol. 2022;13: [eLocator: 857308].

Schaub FX, Dhankani V, Berger AC, et al. Pan‐cancer alterations of the MYC oncogene and its proximal network across the cancer genome Atlas. Cell Syst. 2018;6(3):282‐300.

Hutter C, Zenklusen JC. The cancer genome Atlas: creating lasting value beyond its data. Cell. 2018;173(2):283‐285.