Learning nonparametric ordinary differential equations from noisy data.

Learning nonparametric systems of Ordinary Differential Equations (ODEs) x˙=f(t,x) from noisy data is an emerging machine learning topic. We use the well-developed theory of Reproducing Kernel Hilbert Spaces (RKHS) to define candidates for f for which the solution of the ODE exists and is unique. Learning f consists of solving a constrained optimization problem in an RKHS. We propose a penalty method that iteratively uses the Representer theorem and Euler approximations to provide a numerical solution. We prove a generalization bound for the L2 distance between x and its estimator. Experiments are provided for the FitzHugh-Nagumo oscillator, the Lorenz system, and for predicting the Amyloid level in the cortex of aging subjects. In all cases, we show competitive results compared with the state-of-the-art.

Machine learning to detect the SINEs of cancer.

We previously described an approach called RealSeqS to evaluate aneuploidy in plasma cell-free DNA through the amplification of ~350,000 repeated elements with a single primer. We hypothesized that an unbiased evaluation of the large amount of sequencing data obtained with RealSeqS might reveal other differences between plasma samples from patients with and without cancer. This hypothesis was tested through the development of a machine learning approach called Alu Profile Learning Using Sequencing (A-PLUS) and its application to 7615 samples from 5178 individuals, 2073 with solid cancer and the remainder without cancer. Samples from patients with cancer and controls were prespecified into four cohorts used for model training, analyte integration, and threshold determination, validation, and reproducibility. A-PLUS alone provided a sensitivity of 40.5% across 11 different cancer types in the validation cohort, at a specificity of 98.5%. Combining A-PLUS with aneuploidy and eight common protein biomarkers detected 51% of the cancers at 98.9% specificity. We found that part of the power of A-PLUS could be ascribed to a single feature-the global reduction of AluS subfamily elements in the circulating DNA of patients with solid cancer. We confirmed this reduction through the analysis of another independent dataset obtained with a different approach (whole-genome sequencing). The evaluation of Alu elements may therefore have the potential to enhance the performance of several methods designed for the earlier detection of cancer.

The Origin of Highly Elevated Cell-Free DNA in Healthy Individuals and Patients with Pancreatic, Colorectal, Lung, or Ovarian Cancer.

Cell-free DNA (cfDNA) concentrations from patients with cancer are often elevated compared with those of healthy controls, but the sources of this extra cfDNA have never been determined. To address this issue, we assessed cfDNA methylation patterns in 178 patients with cancers of the colon, pancreas, lung, or ovary and 64 patients without cancer. Eighty-three of these individuals had cfDNA concentrations much greater than those generally observed in healthy subjects. The major contributor of cfDNA in all samples was leukocytes, accounting for ∼76% of cfDNA, with neutrophils predominating. This was true regardless of whether the samples were derived from patients with cancer or the total plasma cfDNA concentration. High levels of cfDNA observed in patients with cancer did not come from either neoplastic cells or surrounding normal epithelial cells from the tumor's tissue of origin. These data suggest that cancers may have a systemic effect on cell turnover or DNA clearance. SIGNIFICANCE: The origin of excess cfDNA in patients with cancer is unknown. Using cfDNA methylation patterns, we determined that neither the tumor nor the surrounding normal tissue contributes this excess cfDNA-rather it comes from leukocytes. This finding suggests that cancers have a systemic impact on cell turnover or DNA clearance. See related commentary by Thierry and Pisareva, p. 2122. This article is featured in Selected Articles from This Issue, p. 2109.

Functional characterization of all CDKN2A missense variants and comparison to in silico models of pathogenicity.

Interpretation of variants identified during genetic testing is a significant clinical challenge. In this study, we developed a high-throughput CDKN2A functional assay and characterized all possible CDKN2A missense variants. We found that 40% of all missense variants were functionally deleterious. We also used our functional classification to assess the performance of in silico models that predict the effect of variants, including recently reported models based on machine learning. Notably, we found that all in silico models similarly when compared to our functional classifications with accuracies of 54.6 - 70.9%. Furthermore, while we found that functionally deleterious variants were enriched within ankyrin repeats, rarely were all missense variants at a single residue functionally deleterious. Our functional classifications are a resource to aid the interpretation of CDKN2A variants and have important implications for the application of variant interpretation guidelines, particularly the use of in silico models for clinical variant interpretation.

Circulating Tumor DNA Analysis Guiding Adjuvant Therapy in Stage II Colon Cancer.

BACKGROUND: The role of adjuvant chemotherapy in stage II colon cancer continues to be debated. The presence of circulating tumor DNA (ctDNA) after surgery predicts very poor recurrence-free survival, whereas its absence predicts a low risk of recurrence. The benefit of adjuvant chemotherapy for ctDNA-positive patients is not well understood. METHODS: We conducted a trial to assess whether a ctDNA-guided approach could reduce the use of adjuvant chemotherapy without compromising recurrence risk. Patients with stage II colon cancer were randomly assigned in a 2:1 ratio to have treatment decisions guided by either ctDNA results or standard clinicopathological features. For ctDNA-guided management, a ctDNA-positive result at 4 or 7 weeks after surgery prompted oxaliplatin-based or fluoropyrimidine chemotherapy. Patients who were ctDNA-negative were not treated. The primary efficacy end point was recurrence-free survival at 2 years. A key secondary end point was adjuvant chemotherapy use. RESULTS: Of the 455 patients who underwent randomization, 302 were assigned to ctDNA-guided management and 153 to standard management. The median follow-up was 37 months. A lower percentage of patients in the ctDNA-guided group than in the standard-management group received adjuvant chemotherapy (15% vs. 28%; relative risk, 1.82; 95% confidence interval [CI], 1.25 to 2.65). In the evaluation of 2-year recurrence-free survival, ctDNA-guided management was noninferior to standard management (93.5% and 92.4%, respectively; absolute difference, 1.1 percentage points; 95% CI, -4.1 to 6.2 [noninferiority margin, -8.5 percentage points]). Three-year recurrence-free survival was 86.4% among ctDNA-positive patients who received adjuvant chemotherapy and 92.5% among ctDNA-negative patients who did not. CONCLUSIONS: A ctDNA-guided approach to the treatment of stage II colon cancer reduced adjuvant chemotherapy use without compromising recurrence-free survival. (Supported by the Australian National Health and Medical Research Council and others; DYNAMIC Australian New Zealand Clinical Trials Registry number, ACTRN12615000381583.).

Supervised mutational signatures for obesity and other tissue-specific etiological factors in cancer.

Determining the etiologic basis of the mutations that are responsible for cancer is one of the fundamental challenges in modern cancer research. Different mutational processes induce different types of DNA mutations, providing 'mutational signatures' that have led to key insights into cancer etiology. The most widely used signatures for assessing genomic data are based on unsupervised patterns that are then retrospectively correlated with certain features of cancer. We show here that supervised machine-learning techniques can identify signatures, called SuperSigs, that are more predictive than those currently available. Surprisingly, we found that aging yields different SuperSigs in different tissues, and the same is true for environmental exposures. We were able to discover SuperSigs associated with obesity, the most important lifestyle factor contributing to cancer in Western populations.

Feasibility of blood testing combined with PET-CT to screen for cancer and guide intervention.

Cancer treatments are often more successful when the disease is detected early. We evaluated the feasibility and safety of multicancer blood testing coupled with positron emission tomography-computed tomography (PET-CT) imaging to detect cancer in a prospective, interventional study of 10,006 women not previously known to have cancer. Positive blood tests were independently confirmed by a diagnostic PET-CT, which also localized the cancer. Twenty-six cancers were detected by blood testing. Of these, 15 underwent PET-CT imaging and nine (60%) were surgically excised. Twenty-four additional cancers were detected by standard-of-care screening and 46 by neither approach. One percent of participants underwent PET-CT imaging based on false-positive blood tests, and 0.22% underwent a futile invasive diagnostic procedure. These data demonstrate that multicancer blood testing combined with PET-CT can be safely incorporated into routine clinical care, in some cases leading to surgery with intent to cure.

Revisiting the tumorigenesis timeline with a data-driven generative model.

Cancer is driven by the sequential accumulation of genetic and epigenetic changes in oncogenes and tumor suppressor genes. The timing of these events is not well understood. Moreover, it is currently unknown why the same driver gene change appears as an early event in some cancer types and as a later event, or not at all, in others. These questions have become even more topical with the recent progress brought by genome-wide sequencing studies of cancer. Focusing on mutational events, we provide a mathematical model of the full process of tumor evolution that includes different types of fitness advantages for driver genes and carrying-capacity considerations. The model is able to recapitulate a substantial proportion of the observed cancer incidence in several cancer types (colorectal, pancreatic, and leukemia) and inherited conditions (Lynch and familial adenomatous polyposis), by changing only 2 tissue-specific parameters: the number of stem cells in a tissue and its cell division frequency. The model sheds light on the evolutionary dynamics of cancer by suggesting a generalized early onset of tumorigenesis followed by slow mutational waves, in contrast to previous conclusions. Formulas and estimates are provided for the fitness increases induced by driver mutations, often much larger than previously described, and highly tissue dependent. Our results suggest a mechanistic explanation for why the selective fitness advantage introduced by specific driver genes is tissue dependent.