IntroductionA common task in the analysis of microbial communities involves assigning taxonomic labels to the sequences derived from organisms found in the communities. Frequently, such labels are assigned using machine learning algorithms that are trained to recognize individual taxonomic groups based on training data sets that comprise sequences with known taxonomic labels. Ideally, the training data should rely on labels that are experimentally verified—formal taxonomic labels require knowledge of physical and biochemical properties of organisms that cannot be directly inferred from sequence alone. However, the labels associated with sequences in biological databases are most commonly computational predictions which themselves may rely on computationally-generated data—a process commonly referred to as “transitive annotation.”MethodsIn this manuscript we explore the implications of training a machine learning classifier (the Ribosomal Database Project’s Bayesian classifier in our case) on data that itself has been computationally generated. We generate new training examples based on 16S rRNA data from a metagenomic experiment, and evaluate the extent to which the taxonomic labels predicted by the classifier change after re-training.ResultsWe demonstrate that even a few computationally-generated training data points can significantly skew the output of the classifier to the point where entire regions of the taxonomic space can be disturbed.Discussion and conclusionsWe conclude with a discussion of key factors that affect the resilience of classifiers to transitively-annotated training data, and propose best practices to avoid the artifacts described in our paper.

Lead is a poisonous metal that our bodies cannot use. Lead poisoning can cause learning, hearing, and behavioral problems, and can harm your child’s brain, kidneys, and other organs. Lead in the body stops good minerals such as iron and calcium from working right. Some of these effects may be permanent.

The file contains the first time tests of coagulation, liver, kidney indices, deceased group (1) and survival group (2). (XLSX)

We evaluated the effect of puerarin on spatial learning and memory ability of mice with chronic alcohol poisoning. A total of 30 male C57BL/6 mice were randomly divided into model, puerarin, and control groups (n=10 each). The model group received 60% (v/v) ethanol by intragastric administration followed by intraperitoneal injection of normal saline 30 min later. The puerarin group received intragastric 60% ethanol followed by intraperitoneal puerarin 30 min later, and the control group received intragastric saline followed by intraperitoneal saline. Six weeks after treatment, the Morris water maze and Tru Scan behavioral tests and immunofluorescence staining of cerebral cortex and hippocampal neurons (by Neu-N) and microglia (by Ib1) were conducted. Glutamic acid (Glu) and gamma amino butyric acid (GABA) in the cortex and hippocampus were assayed by high-performance liquid chromatography (HPLC), and tumor necrosis factor (TNF)-α and interleukin (IL)-1β were determined by ELISA. Compared with mice in the control group, escape latency and distance were prolonged, and spontaneous movement distance was shortened (P

The dataset consists of the annual number of deaths of elephants due to unnatural deaths by causes like poaching, electrocution, train accidents, and poisoning across states.

Statistics concerning different logistic regression models’ abilities to predict opioid poisoning calls to the APCCa in US dogs (2005–2014).

Number of coefficients in models fitted using various logistic regression models examining the associations between dog-level variables and a poisoning call to the APCCa being related to cannabinoids or opioids (2005–2014).

The impact of label noise on the performance of ranking models.