A detailed comparison of analysis processes for MCC-IMS data in disease classification—Automated methods can replace manual peak annotations

Abstract

MotivationDisease classification from molecular measurements typically requires an analysis pipeline from raw noisy measurements to final classification results. Multi capillary column—ion mobility spectrometry (MCC-IMS) is a promising technology for the detection of volatile organic compounds in the air of exhaled breath. From raw measurements, the peak regions representing the compounds have to be identified, quantified, and clustered across different experiments. Currently, several steps of this analysis process require manual intervention of human experts. Our goal is to identify a fully automatic pipeline that yields competitive disease classification results compared to an established but subjective and tedious semi-manual process.MethodWe combine a large number of modern methods for peak detection, peak clustering, and multivariate classification into analysis pipelines for raw MCC-IMS data. We evaluate all combinations on three different real datasets in an unbiased cross-validation setting. We determine which specific algorithmic combinations lead to high AUC values in disease classifications across the different medical application scenarios.ResultsThe best fully automated analysis process achieves even better classification results than the established manual process. The best algorithms for the three analysis steps are (i) SGLTR (Savitzky-Golay Laplace-operator filter thresholding regions) and LM (Local Maxima) for automated peak identification, (ii) EM clustering (Expectation Maximization) and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) for the clustering step and (iii) RF (Random Forest) for multivariate classification. Thus, automated methods can replace the manual steps in the analysis process to enable an unbiased high throughput use of the technology.

BALSAM—An Interactive Online Platform for Breath Analysis, Visualization and Classification

“…Two approaches, Probe-Clustering and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) [ 45 ] are available in the platform (for a direct comparison see Figure S1 ). Probe-Clustering is closely related to the “Grid-Squares” approach termed by Horsch et al [ 24 ] and makes use of a RT scaling method similar to the implementation in Visual Now [ 44 ]. Its application is vital—the peak is assigned to specific peaks and needs to be deterministic and should not differ between runs, as we rely on peaks as a proxy for metabolites during classification.…”

“…In previous years, finding optimal combinations of pre-processing and analysis techniques that facilitate and reduce the need for manual interpretation of MCC-IMS data could not be fully solved [ 23 , 24 , 25 ]. While several techniques and tools have been established to facilitate these analyses, their use remains impractical for daily clinical practice for biomedical researchers who often lack the necessary time or are unfamiliar with a programming environment.…”

“…While providing a variety of methods to the community, it lacks support for MCC-IMS data. With BALSAM (Breath AnaLysis viSualizAtion Metabolite discovery), we present an accessible web-platform that integrates state-of-the-art preprocessing and analysis techniques (see [ 24 , 25 , 31 ]). BALSAM combines automatic cross-validation for reproducible machine learning, model performance estimation and an easily interpretable classification strategy to guide the biomedical decision-making process.…”

“…Additionally, it does not require a background in computer science or statistics and greatly reduces the manual intervention required. We follow recommended techniques and provide capabilities for MCC-IMS preprocessing, peak detection and peak alignment [ 24 , 25 , 31 ]. Our main focus is the analysis of data from MCC-IMS devices.…”

Full Workflows for the Analysis of Gas Chromatography—Ion Mobility Spectrometry in Foodomics: Application to the Analysis of Iberian Ham Aroma

“…From a more practical perspective, manipulating high-dimensional data can lead to time-consuming, computationally intractable analysis problems. For this reason, Horsch et al [ 45 ] systematically compared both manual and automated strategies for GC-IMS data analysis.…”