Ranks underlie outcome of combining classifiers: Quantitative roles for diversity and accuracy

A multivariate cell-based assay for blood-based diagnostics enhances lung cancer risk stratification

The indicator cell assay platform (iCAP) is a tool for blood-based diagnostics that addresses the low signal-to-noise ratio of blood biomarkers by using cells as biosensors. The assay exposes small volumes of patient serum to standardized cells in culture and classifies disease by machine learning analysis of the gene expression readout from the cells. We developed the lung cancer iCAP (LC-iCAP) as a rule-out test for nodule management in computed tomography (CT)-based lung-cancer screening. We performed analytical optimization, rigorous reproducibility testing, and assessed performance in a study with prospective-specimen-collection, retrospective-blinded-evaluation (PRoBE) design. LC-iCAP achieved an AUC of 0.64 (95% CI, 0.51-0.76) on the ROC curve in validation. Post-validation integration of the assay readout with CT-based features showed improved clinical utility compared to the Mayo Clinic model, with 90% sensitivity, 64% specificity, and 95% negative predictive value at 25% prevalence. The lung-cancer specific readout was enriched for hypoxia-responsive genes and was reproducible across different indicator cell lineages. This is the first validation study of an iCAP and the first application for early cancer detection. The LC-iCAP uses immortalized cells, is scalable and cost-effective and has a multivariate readout. This study supports its potential as a next-generation multivalent platform for precision medicine applications in multi-cancer screening and drug development.

Key Points

We developed the LC-iCAP, novel approach for liquid biopsies that uses cultured cells as biosensors. The cells detect cancer signals in serum and transduce them into standardized gene expression profiles, which are analyzed by machine learning for disease classification. The assay is inexpensive and scalable and has a multivariate readout with potential utility for precision medicine and multi-cancer early detection.A LC-iCAP-based lung cancer risk classifier demonstrated improved specificity compared to existing tests, suggesting meaningful clinical utility for managing indeterminate pulmonary nodules.We identified a lung-cancer specific transcriptional response to hypoxia in the assay readout, implicating HIF1A and HIF2A activity in the response consistent with known lung cancer biology and highlighting the platform's mechanistic relevance.Standardized controls and validation studies demonstrated assay reproducibility, lineage stability, and detection of technical errors-supporting the platform's readiness for clinical deployment.

The DIRAC framework: Geometric structure underlies roles of diversity and accuracy in combining classifiers

Combining classification systems potentially improves predictive accuracy, but outcomes have proven impossible to predict. Similar to improving binary classification with fusion, fusing ranking systems most commonly increases Pearson or Spearman correlations with a target when the input classifiers are "sufficiently good" (generalized as "accuracy") and "sufficiently different" (generalized as "diversity"), but the individual and joint quantitative influence of these factors on the final outcome remains unknown. We resolve these issues. Building on our previous empirical work establishing the DIRAC (DIversity of Ranks and ACcuracy) framework, which accurately predicts the outcome of fusing binary classifiers, we demonstrate that the DIRAC framework similarly explains the outcome of fusing ranking systems. Specifically, precise geometric representation of diversity and accuracy as angle-based distances within rank-based combinatorial structures (permutahedra) fully captures their synergistic roles in rank approximation, uncouples them from the specific metrics of a given problem, and represents them as generally as possible.

Predicting drug toxicity at the intersection of informatics and biology: DTox builds a foundation

Hao et al. (2022) present DTox (deep learning for toxicology), a neural network designed to predict and probe the sites and potential mechanisms underlying chemical toxicity; results provide a map to facilitate modular testing and improvements across multiple disparate applications.

Improving SDG Classification Precision Using Combinatorial Fusion

Combinatorial fusion algorithm (CFA) is a machine learning and artificial intelligence (ML/AI) framework for combining multiple scoring systems using the rank-score characteristic (RSC) function and cognitive diversity (CD). When measuring the relevance of a publication or document with respect to the 17 Sustainable Development Goals (SDGs) of the United Nations, a classification scheme is used. However, this classification process is a challenging task due to the overlapping goals and contextual differences of those diverse SDGs. In this paper, we use CFA to combine a topic model classifier (Model A) and a semantic link classifier (Model B) to improve the precision of the classification process. We characterize and analyze each of the individual models using the RSC function and CD between Models A and B. We evaluate the classification results from combining the models using a score combination and a rank combination, when compared to the results obtained from human experts. In summary, we demonstrate that the combination of Models A and B can improve classification precision only if these individual models perform well and are diverse.