Systematic evaluation of supervised classifiers for fecal microbiota-based prediction of colorectal cancer

Abstract

Predicting colorectal cancer (CRC) based on fecal microbiota presents a promising method for non-invasive screening of CRC, but the optimization of classification models remains an unaddressed question. The purpose of this study was to systematically evaluate the effectiveness of different supervised machine-learning models in predicting CRC in two independent eastern and western populations. The structures of intestinal microflora in feces in Chinese population (N = 141) were determined by 454 FLX pyrosequencing, and different supervised classifiers were employed to predict CRC based on fecal microbiota operational taxonomic unit (OTUs). As a result, Bayes Net and Random Forest displayed higher accuracies than other algorithms in both populations, although Bayes Net was found with a lower false negative rate than that of Random Forest. Gut microbiota-based prediction was more accurate than the standard fecal occult blood test (FOBT), and the combination of both approaches further improved the prediction accuracy. Moreover, when unclassified OTUs were used as input, the BayesDMNB text algorithm achieved higher accuracy in the Chinese population (AUC=0.994). Taken together, our results suggest that Bayes Net classification model combined with unclassified OTUs may present an accurate method for predicting CRC based on the compositions of gut microbiota.

Synbindin deficiency inhibits colon carcinogenesis by attenuating Wnt cascade and balancing gut microbiome

“…A striking example is our recent identification of the link between Fusobacterium nucleatum and the chemoresistance to CRC . We and others also showed that gut microbiota alone could predict CRC well, highlighting a potential clinical application of gut microbiota as a noninvasive screen strategy for early detection of CRC in population . Despite the enormous efforts and substantial progress, only minor improvements have been achieved in the clinical outcome of patients with CRC.…”

“…We previously reported that gut microbiome was able to noninvasively screen CRC by employing supervised machinelearning algorithms, particularly Bayes Net and Random Forest algorithm. 11 Similarly, we tried to study whether the altered gut micobiome could discriminate tumor-bearing mice from control mice, or more strikingly, distinguish Synbindin flox mice from Synbindin ΔIEC mice. As expected, microbiota composition distinguished tumor-bearing mice from tumor-free mice with the performance of 1 AUC even at the phylum level( Fig.…”

“…10 We and others also showed that gut microbiota alone could predict CRC well, highlighting a potential clinical application of gut microbiota as a noninvasive screen strategy for early detection of CRC in population. 11,12 Despite the enormous efforts and substantial progress, only minor improvements have been achieved in the clinical outcome of patients with CRC. A better understanding of the signaling cascades as well as the interplay between microbiota and host are needed to advance the therapy of this disease.…”

Rise of PD‐L1 expression during metastasis of colorectal cancer: Implications for immunotherapy

“…Colorectal cancer (CRC) is one of the leading cancers worldwide, and its incidence has increased by 22% from 2000 to 2013 in the United States. In China, the prevention and treatment of CRC remain to be highly challenging tasks, given the complicated risk factors such as genetic background, environment, microbiota and immune‐related disorders. In metastatic CRC, the outcomes related to surgery or adjuvant therapy remain poor .…”

Variability of core microbiota in newly diagnosed treatment-naïve paediatric Inflammatory Bowel Disease patients

“…In the present study, acceptably sensitive classification or characterization could also not be achieved by unsupervised approaches, such as clustering and PCoA, while application of a supervised technique like group-regularised logistic regression analysis did achieve high predictive accuracy. This example of supervised machine learning is particularly useful for pattern recognition in highly complex data sets, such as the gut microbiome (40,41). Obviously, as mucosal biopsies are not likely to be routinely harvested in the follow-up of IBD patients, faecal samples have substantial advantages as biomarkers of both disease treatment and activity.…”