Richard Simon

Genomic tools are demonstrating that many human diseases are molecularly heterogeneous and likely to respond differently to molecularly targeted therapeutics. For many widely used treatments, the number of patients needed to treat (NNT) for each patient who benefits is large indicating that many patients are being exposed to the risks of serious adverse effects although they do not benefit from the drug. Consequently, more accurately determining the intended use population for new therapeutics is of increased importance. In this paper, we describe a new paradigm for identifying and internally validating an estimate of the intended use population in randomized phase III clinical trials. The approach preserves the type I error of the trial and approaches determination of the intended use population as a classification problem, not a multiple hypothesis testing problems.

The APOBEC gene family of cytidine deaminases plays important roles in DNA repair and mRNA editing. In many cancers, APOBEC3B increases the mutation load, generating clusters of closely spaced, single-strand-specific DNA substitutions with a characteristic hypermutation signature. Some studies also suggested a possible involvement of APOBEC3A, REV1, UNG, and FHIT in molecular processes affecting APOBEC mutagenesis. It is important to understand how mutagenic processes linked to the activity of these genes may affect sensitivity of cancer cells to treatment. We used information from the Cancer Cell Line Encyclopedia and the Genomics of Drug Sensitivity in Cancer resources to examine associations of the prevalence of APOBEC-like motifs and mutational loads with expression of APOBEC3A, APOBEC3B, REV1, UNG, and FHIT and with cell line chemosensitivity to 255 antitumor drugs. Among the five genes, APOBEC3B expression levels were bimodally distributed, whereas expression of APOBEC3A, REV1, UNG, and FHIT was unimodally distributed. The majority of the cell lines had low levels of APOBEC3A expression. The strongest correlations of gene expression levels with mutational loads or with measures of prevalence of APOBEC-like motif counts and kataegis clusters were observed for REV1, UNG, and APOBEC3A. Sensitivity or resistance of cell lines to JQ1, palbociclib, bicalutamide, 17-AAG, TAE684, MEK inhibitors refametinib, PD-0325901, and trametinib and a number of other agents was correlated with candidate gene expression levels or with abundance of APOBEC-like motif clusters in specific cancers or across cancer types. We observed correlations of expression levels of the five candidate genes in cell line models with sensitivity to cancer drug treatment. We also noted suggestive correlations between measures of abundance of APOBEC-like sequence motifs with drug sensitivity in small samples of cell lines from individual cancer categories, which require further validation in larger datasets. Molecular mechanisms underlying the links between the activities of the products of each of the five genes, the resulting mutagenic processes, and sensitivity to each category of antitumor agents require further investigation.

I provide a commentary on two papers “Probability estimation with machine learning methods for dichotomous and multicategory outcome: Theory” by Jochen Kruppa, Yufeng Liu, Gérard Biau, Michael Kohler, Inke R. König, James D. Malley, and Andreas Ziegler; and “Probability estimation with machine learning methods for dichotomous and multicategory outcome: Applications” by Jochen Kruppa, Yufeng Liu, Hans‐Christian Diener, Theresa Holste, Christian Weimar, Inke R. König, and Andreas Ziegler. Those papers provide an up‐to‐date review of some popular machine learning methods for class probability estimation and compare those methods to logistic regression modeling in real and simulated datasets.

The introduction of next-generation sequencing (NGS) technology has made it possible to detect genomic alterations within tumor cells on a large scale. However, most applications of NGS show the genetic content of mixtures of cells. Recently developed single cell sequencing technology can identify variation within a single cell. Characterization of multiple samples from a tumor using single cell sequencing can potentially provide information on the evolutionary history of that tumor. This may facilitate understanding how key mutations accumulate and evolve in lineages to form a heterogeneous tumor. We provide a computational method to infer an evolutionary mutation tree based on single cell sequencing data. Our approach differs from traditional phylogenetic tree approaches in that our mutation tree directly describes temporal order relationships among mutation sites. Our method also accommodates sequencing errors. Furthermore, we provide a method for estimating the proportion of time from the earliest mutation event of the sample to the most recent common ancestor of the sample of cells. Finally, we discuss current limitations on modeling with single cell sequencing data and possible improvements under those limitations. Inferring the temporal ordering of mutational sites using current single cell sequencing data is a challenge. Our proposed method may help elucidate relationships among key mutations and their role in tumor progression.

Accurate class probability estimation is important for medical decision making but is challenging, particularly when the number of candidate features exceeds the number of cases. Special methods have been developed for nonprobabilistic classification, but relatively little attention has been given to class probability estimation with numerous candidate variables. In this paper, we investigate overfitting in the development of regularized class probability estimators. We investigate the relation between overfitting and accurate class probability estimation in terms of mean square error. Using simulation studies based on real datasets, we found that some degree of overfitting can be desirable for reducing mean square error. We also introduce a mean square error decomposition for class probability estimation that helps clarify the relationship between overfitting and prediction accuracy.

Recent developments in high-throughput genomic technologies make it possible to have a comprehensive view of genomic alterations in tumors on a whole genome scale. Only a small number of somatic alterations detected in tumor genomes are driver alterations which drive tumorigenesis. Most of the somatic alterations are passengers that are neutral to tumor cell selection. Although most research efforts are focused on analyzing driver alterations, the passenger alterations also provide valuable information about the history of tumor development.

High-throughput ?omics? technologies that generate molecular profiles for biospecimens have been extensively used in preclinical studies to reveal molecular subtypes and elucidate the biological mechanisms of disease, and in retrospective studies on clinical specimens to develop mathematical models to predict clinical endpoints. Nevertheless, the translation of these technologies into clinical tests that are useful for guiding management decisions for patients has been relatively slow. It can be difficult to determine when the body of evidence for an omics-based test is sufficiently comprehensive and reliable to support claims that it is ready for clinical use, or even that it is ready for definitive evaluation in a clinical trial in which it may be used to direct patient therapy. Reasons for this difficulty include the exploratory and retrospective nature of many of these studies, the complexity of these assays and their application to clinical specimens, and the many potential pitfalls inherent in the development of mathematical predictor models from the very high-dimensional data generated by these omics technologies. Here we present a checklist of criteria to consider when evaluating the body of evidence supporting the clinical use of a predictor to guide patient therapy. Included are issues pertaining to specimen and assay requirements, the soundness of the process for developing predictor models, expectations regarding clinical study design and conduct, and attention to regulatory, ethical, and legal issues. The proposed checklist should serve as a useful guide to investigators preparing proposals for studies involving the use of omics-based tests. The US National Cancer Institute plans to refer to these guidelines for review of proposals for studies involving omics tests, and it is hoped that other sponsors will adopt the checklist as well.

Developments in biotechnology have stimulated the use of predictive biomarkers to identify patients who are likely to benefit from a targeted therapy. Several randomized phase III designs have been introduced for development of a targeted therapy using a diagnostic test. Most such designs require biomarkers measured before treatment. In many cases, it has been very difficult to identify such biomarkers. Promising candidate biomarkers can sometimes be effectively measured after a short run-in period on the new treatment. We introduce a new design for phase III trials with a candidate predictive pharmacodynamic biomarker measured after a short run-in period. Depending on the therapy and the biomarker performance, the trial would either randomize all patients but perform a separate analysis on the biomarker-positive patients or only randomize marker-positive patients after the run-in period. We evaluate the proposed design compared with the conventional phase III design and discuss how to design a run-in trial based on phase II studies. The proposed design achieves a major sample size reduction compared with the conventional randomized phase III design in many cases when the biomarker has good sensitivity (≥0.7) and specificity (≥0.7). This requires that the biomarker be measured accurately and be indicative of drug activity. However, the proposed design loses some of its advantage when the proportion of potential responders is large (>50%) or the effect on survival from run-in period is substantial. Incorporating a pharmacodynamic biomarker requires careful consideration but can expand the capacity of clinical trials to personalize treatment decisions and enhance therapeutics development.

Identification of genes that are synthetic lethal to p53 is an important strategy for anticancer therapy as p53 mutations have been reported to occur in more than half of all human cancer cases. Although genome-wide RNAi screening is an effective approach to finding synthetic lethal genes, it is costly and labor-intensive.