Paclitaxel Response Can Be Predicted With Interpretable Multi-Variate Classifiers Exploiting DNA-Methylation and miRNA Data

Abstract

To address the problem of resistance to paclitaxel treatment, we have investigated to which extent is possible to predict Breast Cancer (BC) patient response to this drug. We carried out a large-scale tumor-based prediction analysis using data from the US National Cancer Institute’s Genomic Data Commons. These data sets comprise the responses of BC patients to paclitaxel along with six molecular profiles of their tumors. We assessed 10 Machine Learning (ML) algorithms on each of these profiles and evaluated the resulting 60 classifiers on the same BC patients. DNA methylation and miRNA profiles were the most informative overall. In combination with these two profiles, ML algorithms selecting the smallest subset of molecular features generated the most predictive classifiers: a complexity-optimized XGBoost classifier based on CpG island methylation extracted a subset of molecular factors relevant to predict paclitaxel response (AUC = 0.74). A CpG site methylation-based Decision Tree (DT) combining only 2 of the 22,941 considered CpG sites (AUC = 0.89) and a miRNA expression-based DT employing just 4 of the 337 analyzed mature miRNAs (AUC = 0.72) reveal the molecular types associated to paclitaxel-sensitive and resistant BC tumors. A literature review shows that features selected by these three classifiers have been individually linked to the cytotoxic-drug sensitivities and prognosis of BC patients. Our work leads to several molecular signatures, unearthed from methylome and miRNome, able to anticipate to some extent which BC tumors respond or not to paclitaxel. These results may provide insights to optimize paclitaxel-therapies in clinical practice.

Concise Polygenic Models for Cancer-Specific Identification of Drug-Sensitive Tumors from Their Multi-Omics Profiles

“…Given their suitability for high-throughput experiments [49], they are also the model for which most phenotypic and pharmacological information is publicly available. For many drugs, only much smaller amounts of pharmacogenomics datasets, if any, has been released using more patient-relevant in vivo models [59][60][61][62]. Consequently, often the only data that can be used to predict drug sensitivity comes from cell lines, thus having their niche in guiding precision oncology efforts [55].…”

Machine‐learning scoring functions for structure‐based virtual screening

“…11,18 Third, the first prospective results of a CNN-based SF are not better in any way to those using more established ML algorithms (Table 4). Last, as any other ML algorithm, DL algorithms are not optimal for many supervised learning problems 17,[19][20][21][22] and SBVS could easily be another one of them.…”

“…17 We will not use the term AI in this review because it conveys the notion that a categorical change has occurred with respect to ML, which is not the case in the research area of SF development, 11,18 as well as in many others. 17,[19][20][21][22] The term AI is, however, appropriate as a reminder of aspirations and achievements. ML-based SFs reached long time ago the level where it is able to identify new potent actives better than a human expert would do using the same training data.…”

“…As a result, the captivating and easy‐to‐explain parallelism between artificial neural networks and human neural networks (AI vs. human intelligence) is being promoted again 17 . We will not use the term AI in this review because it conveys the notion that a categorical change has occurred with respect to ML, which is not the case in the research area of SF development, 11,18 as well as in many others 17,19–22 …”

Current Perspectives on Taxanes: Focus on Their Bioactivity, Delivery and Combination Therapy

“…The most recent study discussing their role in cell resistance to taxanes was performed by Chen et al [ 113 ], who identified two miRNAs, miR-335-5p and hsa-let-7c-5p, and their gene targets, chemokine (C-X-C motif) ligand 9, C-C chemokine receptor type 7, and suppressor of cytokine signaling 1, which are all linked to cell resistance to taxanes. Certain miRNA profiles are typical for paclitaxel-sensitive or chemoresistant tumors; these molecules could be used in the initial screening preceding the personalized treatment of cancer patients [ 114 ].…”