Ready to Learn About Your Gut Health with AI?

Colorectal cancer is currently one of the most common cancers worldwide. However, many existing screening methods—such as genetic testing and fecal occult blood tests—can produce false-positive results, making early and accurate diagnosis challenging.

To help address this, a research team from the Department of Chemistry and Materials Science, School of Science, Xi’an Jiaotong-Liverpool University (XJTLU) conducted a systematic study to search for measurable “gas signals” that may reflect gut health, focusing on volatile organic compounds (VOCs) released from fecal samples. The first author of this work is Weiyu Xiao, School of Science Ph.D. student with a research focus on metal oxide-based sensing including its design, development, validation as well as sensing mechanism study.

In this study, the team recruited three groups of participants: patients with colorectal cancer, patients with adenomas (precancerous lesions), and healthy volunteers. In total, Fecal samples were collected from 37 colorectal cancer patients, 44 adenoma patients, and 55 healthy donors. Using gas chromatography–mass spectrometry (GC-MS), Weiyu Xiao and Dr Qiuchen Dong team analyzed these samples and identified 80 volatile organic compounds (VOCs) as potential biomarkers related to gut health.

The results showed clear differences in the levels of certain gases among the three groups. In particular, isopropanol levels differed significantly between colorectal cancer patients, adenoma patients, and healthy individuals. By combining these data with machine-learning models, we were able to accurately predict the gut health status of unknown samples as shown in Figure 1 and Figure 2.

Figure 1. (A) and (B) The predictions of PCA algorithm for three groups and supervised machine learning models are based on fecal isopropanol gas concentration detected by ZnO/IrO_x-based gas sensor and fecal p-cresol analyzed by GC-MS. (C) The predictions of the K-means algorithm for three groups. (D) Supervised machine learning models are based on fecal isopropanol gas concentration detected by ZnO/IrO_x-based gas sensor and fecal p-cresol, 2-methylbutanoic acid and 3-methylbutanoic acid analyzed by GC-MS.

Figure 2. (A) Electrodeposited 40 min ZnO including current curves toward different feal isopropanol gas released from human feces with condition of CRC patients, polyps and healthy donors at 40  and the humidity of RH 15 %. (B) The quantitative analysis of fecal isopropanol gas. (C) Box plot and distribution of fecal isopropanol gas from three groups detected by ZnO-based gas sensor. (D) PCA algorithm for three groups based on fecal isopropanol gas concentration detected by ZnO-based gas sensor and fecal p-cresol analyzed by GC-MS.

Overall, the analysis suggests that isopropanol and para-cresol are among the most promising VOCs for the early detection of colorectal cancer. 2-methylbutanoic acid and 3-methylbutanoic acid are two of the supplemental volatile organic compounds that can be used to fine-tune the prediction model due to their different abundance in the three controlled groups.

In addition to biomarker discovery, the team also developed two types of gas sensors to detect these compounds. One sensor was based on pure zinc oxide, while the other used a p–n heterojunction made from zinc oxide and iridium oxide nanomaterials. These materials were fabricated using photolithography and deposited onto specially designed concentric interdigitated electrodes. Electrodeposition and heat treatment were used to achieve stable crystalline structures that enhance the oxidation of isopropanol, improving sensor performance.

This study was conducted in collaboration with the First Affiliated Hospital of Soochow University, led by Dr. Songbing He. The related technology has been filed as two Chinese patent applications (Application Nos. 202510501780.6 and 202511684369.3) and two peer-reviewed publications in Sensors and Actuators B: Chemical (https://doi.org/10.1016/j.snb.2025.139347) and Sensors and Actuators Reports (https://doi.org/10.1016/j.snr.2025.100420), published in mid-December 2025.
In the future, the team are pursuing to combine the sample collection, signal processing with disease early diagnosis in to an integrated platform to further improve the testing efficiency and its point-of-care property.

Content and review：Dr Qiuchen Dong
Editor:Luyao