Differential array sensing for cancer cell classification and novelty detection

Abstract

A series of semi-specific peptides reported in the literature to bind various epitopes on cell surfaces were used in a differential sensing array to pattern cell line identity. The peptides were conjugated to thiazole orange to act as both a fluorescence reporter and a DNA intercalator. Fluorescence data for the peptides exposed to cells, with and without exogenous double stranded DNA (dsDNA), led to chemometric fingerprints for eight cancer cell lines. In contrast to the use of structures meant to act in completely non-specific ways, the use of a limited level of specificity generated linear discriminant score plots with high dimensionality, i.e. several principle components carrying significant variance. The arrays were found to correctly classify the cell lines from 60% to 100% depending upon the cell line. Due to the high dimensionality score plots, the identification of cell lines that were not part of the training set was examined. Support vector machines were used as a novelty detection routine and showed that a cancer line not part of the original training set could be correctly identified as being novel.

Gold nanoparticles‐mediated fluorescent chemical nose sensor for pathogenic diagnosis and phenotype

“…The algorithm could classify an unknown sample to a known group through simple linear fitting, just like the LDA (Linear Discriminant Analysis) used in this experiment. 38,39 Via Fisher multi-class LDA of the fluorescence reaction intensity from GFCEs, we could obtain corresponding two sets of scores, which can be used to map fluorescence fingerprint by Origin 9.0. The discriminant scores were received from SPSS analysis on fluorescence reaction intensity with 95% confidence ellipses for each bacteria.…”

Array‐basierte Sensorik mit der “chemischen Nase” in der Diagnostik und Wirkstoffentdeckung

“…B. der Linear Discriminant Analysis/linearen Diskriminanzanalyse; LDA) wird – basierend auf einfachen, linearen Angleichungen – eine unbekannte Probe in eine der bekannten Gruppen klassifiziert. Mithilfe weiter fortgeschrittener überwachter Algorithmen können komplexere Unterscheidungsmuster generiert werden, die nicht nur unbekannte Proben identifizieren, die eine Übereinstimmung mit einer der ursprünglich trainierten Datenklassen ergeben, sondern auch solche, die zwar ähnlich, aber dennoch unterschiedlich zu den ursprünglichen Klassen sind, oder auch Proben, die deutlich außerhalb jeder bekannten Klasse liegen . Beispiele für derartige Algorithmen umfassen Support Vector Machines (SVM) oder künstliche neuronale Netze (Artificial Neural Networks, ANN); eine detaillierte Diskussion dieser Techniken des maschinellen Lernens sprengt jedoch den Rahmen dieses Kurzaufsatzes.…”

“…Multivariate Analysen der Array‐Daten ergaben, dass das Array acht Krebszelllinien aus verschiedenen Körperteilen unterscheiden konnte. Am interessantesten ist, dass bei Anwendung einer Support Vector Machine (SVM) auf die Daten eine neunte Zelllinie entdeckt und differenziert werden konnte, die nicht in der Testreihe enthalten war …”

More Than a Light Switch: Engineering Unconventional Fluorescent Configurations for Biological Sensing

“…126 Anslyn and coworkers have also discriminated between cancer cell lines with a DNA-fluorophore sensor and introduced the use of a support-vector machine analysis for going beyond simple classification and detecting samples that do not fall into the training data set. 127 Rotello and coworkers have also probed the method of action of drugs on various cancers, with successful prediction of unknown reaction mechanisms through advanced HCA and LDA analysis, as shown in Figure 6B. 116 Looking forward, new arrays are starting to emerge that can detect disease states more broadly (e.g.…”