# Shared Plasma Fatty Acid Profiles in Four Cancer Types Enable Diagnosis and Discrimination of Gastrointestinal and Lung Cancers

## TL;DR

This study shows that blood fatty acid patterns can help diagnose and distinguish between different types of cancer, including gastrointestinal and lung cancers.

## Contribution

The study identifies shared and cancer-specific fatty acid profiles that enable accurate diagnosis and classification of four major cancer types.

## Key findings

- Shared fatty acid alterations were observed across four cancer types, indicating a common metabolic shift.
- Gastrointestinal cancers showed a distinct fatty acid pattern, while lung cancer showed opposite trends.
- PLS-DA models achieved high diagnostic accuracy, with AUC values up to 0.926 for gastric cancer.

## Abstract

Background: Cancer is a leading cause of mortality worldwide, characterized by metabolic reprogramming, including alterations in fatty acid (FA) metabolism. Plasma FA profiles hold promise as non-invasive biomarkers for the diagnosis and classification of cancer. Objectives: This study aimed to investigate the diagnostic potential of plasma FA profiles across four major cancers and to identify shared and cancer-type-specific metabolic alterations. Methods: We examine comprehensive FA profiling of plasma samples from 368 individuals, including patients with colorectal (CRC, n = 94), gastric (GC, n = 55), esophageal (EC, n = 53), and lung cancer (LC, n = 73), alongside 93 healthy controls (HCs) by gas chromatography–mass spectrometry. Data were analyzed using univariate statistics and multivariate modeling analysis. Results: Univariate analysis showed a shared set of altered FAs across the cancer types, demonstrating a shared pan-cancer metabolic shift. A comprehensive comparison revealed a remarkable shared pattern within the gastrointestinal (GI) cancers (GC, CRC, EC), while LC showed opposite trends for most FAs. Partial Least Squares Discriminant Analysis (PLS-DA) models on a 70% training set excellently discriminated each cancer type from HCs. The cross-validation of the model demonstrated robust internal performance with Q2 = 0.675 (LC), 0.559 (GC), 0.774 (CRC), and 0.628 (EC). This is followed by assessing the diagnostic accuracy on a 30% hold-out test set, with area under the curve (AUC) values of 0.686 (LC), 0.926 (GC), 0.905 (CRC), and 0.843 (EC). Conclusions: Plasma FA profiles may provide a potential source of biomarkers, capturing both shared cancer markers and distinct tissue-specific metabolic alterations. These findings highlight the high diagnostic and classificatory potential of FAs alterations in oncology.