Browse
Submit Data
Databases
API
Help

Dataset Information

16 Views

0 Connections

0 Citations

0 Reanalyses

0 Downloads

Omics score: 0

Clustering: how much bias do we need?

ABSTRACT: Scientific investigations in medicine and beyond increasingly require observations to be described by more features than can be simultaneously visualized. Simply reducing the dimensionality by projections destroys essential relationships in the data. Similarly, traditional clustering algorithms introduce data bias that prevents detection of natural structures expected from generic nonlinear processes. We examine how these problems can best be addressed, where in particular we focus on two recent clustering approaches, Phenograph and Hebbian learning clustering, applied to synthetic and natural data examples. Our results reveal that already for very basic questions, minimizing clustering bias is essential, but that results can benefit further from biased post-processing.This article is part of the themed issue 'Mathematical methods in medicine: neuroscience, cardiology and pathology'.

SUBMITTER: Lorimer T

PROVIDER: S-EPMC5434083 | biostudies-literature | 2017 Jun

REPOSITORIES: biostudies-literature

ACCESS DATA

Json Xml

Publications

Clustering: how much bias do we need?

Lorimer Tom T Held Jenny J Stoop Ruedi R

Philosophical transactions. Series A, Mathematical, physical, and engineering sciences 20170601 2096

Scientific investigations in medicine and beyond increasingly require observations to be described by more features than can be simultaneously visualized. Simply reducing the dimensionality by projections destroys essential relationships in the data. Similarly, traditional clustering algorithms introduce data bias that prevents detection of natural structures expected from generic nonlinear processes. We examine how these problems can best be addressed, where in particular we focus on two recent ...[more]

PMID: 28507238

Similar Datasets

Too much information: Why CDCL solvers need to forget learned clauses.

Project description:Conflict-driven clause learning (CDCL) is a remarkably successful paradigm for solving the satisfiability problem of propositional logic. Instead of a simple depth-first backtracking approach, this kind of solver learns the reason behind occurring conflicts in the form of additional clauses. However, despite the enormous success of CDCL solvers, there is still only a limited understanding of what influences the performance of these solvers in what way. Considering different measures, this paper demonstrates, quite surprisingly, that clause learning (without being able to get rid of some clauses) can not only help the solver but can oftentimes deteriorate the solution process dramatically. By conducting extensive empirical analysis, we furthermore find that the runtime distributions of CDCL solvers are multimodal. This multimodality can be seen as a reason for the deterioration phenomenon described above. Simultaneously, it also gives an indication of why clause learning in combination with clause deletion is virtually the de facto standard of SAT solving, in spite of this phenomenon. As a final contribution, we show that Weibull mixture distributions can accurately describe the multimodal distributions. Thus, adding new clauses to a base instance has an inherent effect of making runtimes long-tailed. This insight provides an explanation as to why the technique of forgetting clauses is useful in CDCL solvers apart from the optimization of unit propagation speed.

| S-EPMC9417043 | biostudies-literature

Bias-adjusted spectral clustering in multi-layer stochastic block models.

Project description:We consider the problem of estimating common community structures in multi-layer stochastic block models, where each single layer may not have sufficient signal strength to recover the full community structure. In order to efficiently aggregate signal across different layers, we argue that the sum-of-squared adjacency matrices contain sufficient signal even when individual layers are very sparse. Our method uses a bias-removal step that is necessary when the squared noise matrices may overwhelm the signal in the very sparse regime. The analysis of our method relies on several novel tail probability bounds for matrix linear combinations with matrix-valued coefficients and matrix-valued quadratic forms, which may be of independent interest. The performance of our method and the necessity of bias removal is demonstrated in synthetic data and in microarray analysis about gene co-expression networks.

| S-EPMC10963943 | biostudies-literature

Clustering of classical swine fever virus isolates by codon pair bias.

Project description:BackgroundThe genetic code consists of non-random usage of synonymous codons for the same amino acids, termed codon bias or codon usage. Codon juxtaposition is also non-random, referred to as codon context bias or codon pair bias. The codon and codon pair bias vary among different organisms, as well as with viruses. Reasons for these differences are not completely understood. For classical swine fever virus (CSFV), it was suggested that the synonymous codon usage does not significantly influence virulence, but the relationship between variations in codon pair usage and CSFV virulence is unknown. Virulence can be related to the fitness of a virus: Differences in codon pair usage influence genome translation efficiency, which may in turn relate to the fitness of a virus. Accordingly, the potential of the codon pair bias for clustering CSFV isolates into classes of different virulence was investigated.ResultsThe complete genomic sequences encoding the viral polyprotein of 52 different CSFV isolates were analyzed. This included 49 sequences from the GenBank database (NCBI) and three newly sequenced genomes. The codon usage did not differ among isolates of different virulence or genotype. In contrast, a clustering of isolates based on their codon pair bias was observed, clearly discriminating highly virulent isolates and vaccine strains on one side from moderately virulent strains on the other side. However, phylogenetic trees based on the codon pair bias and on the primary nucleotide sequence resulted in a very similar genotype distribution.ConclusionClustering of CSFV genomes based on their codon pair bias correlate with the genotype rather than with the virulence of the isolates.

| S-EPMC3341591 | biostudies-literature

Neutralizing antibodies against HIV-1: can we elicit them with vaccines and how much do we need?

Project description:Purpose of reviewThis review describes some of the major obstacles that have impeded progress in the development of an effective neutralizing antibody-based HIV-1 vaccine and explains why it may be possible to overcome these obstacles. A renewed interest in the B-cell response in HIV-1-infected individuals is emphasized.Recent findingsNew assay technologies and access to large numbers of clinical specimens have permitted a detailed assessment of the neutralizing antibody response in HIV-1-infected individuals. Recent studies have demonstrated that B cells can be stimulated to generate high titers of broadly cross-reactive neutralizing antibodies against multiple genetic subtypes of the virus. Preliminary evidence suggests that some of these antibodies are directed against epitopes in the CD4 binding site on monomeric gp120, whereas many others are directed against epitopes that remain to be identified.SummaryThe rationale for pursuing an effective neutralizing antibody-based HIV-1 vaccine is strengthened by the recent demonstration of potent neutralizing antibody responses in a subset of HIV-1-infected individuals. Information on how this response develops and what epitopes are targeted could provide the insights that are needed to design improved vaccine strategies.

| S-EPMC2879149 | biostudies-literature

Accumulation Bias in meta-analysis: the need to consider time in error control.

Project description:Studies accumulate over time and meta-analyses are mainly retrospective. These two characteristics introduce dependencies between the analysis time, at which a series of studies is up for meta-analysis, and results within the series. Dependencies introduce bias - Accumulation Bias - and invalidate the sampling distribution assumed for p-value tests, thus inflating type-I errors. But dependencies are also inevitable, since for science to accumulate efficiently, new research needs to be informed by past results. Here, we investigate various ways in which time influences error control in meta-analysis testing. We introduce an Accumulation Bias Framework that allows us to model a wide variety of practically occurring dependencies including study series accumulation, meta-analysis timing, and approaches to multiple testing in living systematic reviews. The strength of this framework is that it shows how all dependencies affect p-value-based tests in a similar manner. This leads to two main conclusions. First, Accumulation Bias is inevitable, and even if it can be approximated and accounted for, no valid p-value tests can be constructed. Second, tests based on likelihood ratios withstand Accumulation Bias: they provide bounds on error probabilities that remain valid despite the bias. We leave the reader with a choice between two proposals to consider time in error control: either treat individual (primary) studies and meta-analyses as two separate worlds - each with their own timing - or integrate individual studies in the meta-analysis world. Taking up likelihood ratios in either approach allows for valid tests that relate well to the accumulating nature of scientific knowledge. Likelihood ratios can be interpreted as betting profits, earned in previous studies and invested in new ones, while the meta-analyst is allowed to cash out at any time and advice against future studies.

| S-EPMC6808047 | biostudies-literature

Need for improvement of cardiovascular health: a clustering method to identify cardiovascular health profiles.

Project description:BackgroundIn France, the overall trend in the incidence of cardiovascular disease is unfavourable, especially in young subjects. This highlights the need to promote cardiovascular health by targeting the main risk factors. Social marketing campaigns to improve cardiovascular health should identify unhealthy behaviour and understand the target audience. The objective of this study was to identify poor cardiovascular health profiles in the French population using a clustering method.MethodsSubjects aged 18-74 years with no history of cardiovascular disease were included from the Esteban cross-sectional survey (2014-16). To evaluate cardiovascular health, seven items were considered as defined by the American Heart Association: blood glucose, blood cholesterol, blood pressure, body mass index, cigarette smoking, diet and physical activity. Cardiovascular health profiles were identified from these seven items by combining multiple correspondence analysis with hierarchical clustering and partitioning.ResultsA total of 1673 subjects were included in the main analysis. Five cardiovascular health profiles were identified: two profiles corresponded to subjects with poor cardiovascular health (mainly older men with a low socioeconomic status), two to subjects with intermediate cardiovascular health (one mainly comprised of young women with a low socioeconomic status and the other of young subjects with a high socioeconomic status) and one to subjects with good cardiovascular health (mainly older women).ConclusionThis description of cardiovascular health profiles, which led to the identification and characterization of target audiences for future population-based prevention campaigns, should be the starting point for improving cardiovascular health in the French population.

| S-EPMC10393485 | biostudies-literature

Non-crossover gene conversions show strong GC bias and unexpected clustering in humans.

Project description:Although the past decade has seen tremendous progress in our understanding of fine-scale recombination, little is known about non-crossover (NCO) gene conversion. We report the first genome-wide study of NCO events in humans. Using SNP array data from 98 meioses, we identified 103 sites affected by NCO, of which 50/52 were confirmed in sequence data. Overlap with double strand break (DSB) hotspots indicates that most of the events are likely of meiotic origin. We estimate that a site is involved in a NCO at a rate of 5.9 × 10(-6)/bp/generation, consistent with sperm-typing studies, and infer that tract lengths span at least an order of magnitude. Observed NCO events show strong allelic bias at heterozygous AT/GC SNPs, with 68% (58-78%) transmitting GC alleles (p = 5 × 10(-4)). Strikingly, in 4 of 15 regions with resequencing data, multiple disjoint NCO tracts cluster in close proximity (?20-30 kb), a phenomenon not previously seen in mammals.

| S-EPMC4404656 | biostudies-literature

How Much Energy Storage can We Afford? On the Need for a Sunflower Society, Aligning Demand with Renewable Supply.

Project description:The worsening climate crisis impels society to accelerate climate action. The attainable speed of the energy transition is ultimately limited by the available energy to build the replacing renewable infrastructures. Decarbonizing the energy system by replacing dispatchable fossil with variable renewable power requires energy storage to match supply with demand. Current storage technologies are energetically expensive to build and operate, thus the demand for storage shapes the fastest possible transition and the probability to exceed 1.5 °C heating. This study explores and quantifies the effect of demanded storage and its technological progress on the fastest possible transition constrained only by energy. The simulation results using three exemplary storage technologies show that storage substantially delays the transition and increases the probability to exceed 1.5 °C heating. Technological progress, if materialized fast, can reduce energy costs of storage; however, storage demand remains a critical driver for climate risks. Consequently, minimizing storage demand through a supply-driven power system effectively reduces climate risks-a paradigm shift towards a solar-aligned "sunflower society".Supplementary informationThe online version contains supplementary material available at 10.1007/s41247-022-00097-y.

| S-EPMC9113673 | biostudies-literature

A Bothrops jararaca venom subproteome to benchmark a spectral clustering algorithm by controlling the selection bias

Project description:The clustering of mass spectra is a critical component of many proteomics applications. The clustering validation science is just as important, having evolved side by side with the clustering algorithms themselves. In this work, we build on Rieder et al. 's cluster validation framework, and we discuss the problem of selection bias in cluster validation measures; we introduce an assessment measure that is biased toward the number of peptide ion species; we introduce a cluster assessment framework for proteomics and demonstrate its importance by evaluating the performance of 8 clustering algorithms in 7 proteomics datasets, and we discuss the tradeoff between assessment measures. Finally, the validation methods presented here can be of broad applicability be-yond the clustering of mass spectra. This PRIDE entry describes in detail sample preparation, LC-MS/MS analysis, and protein identification of one of the proteomics datasets used in this work (> 10 kDa Bothrops jararaca snake venom proteome).

2022-03-16 | PXD022124 | Pride

Evidence of Experimental Bias in the Life Sciences: Why We Need Blind Data Recording.

Project description:Observer bias and other "experimenter effects" occur when researchers' expectations influence study outcome. These biases are strongest when researchers expect a particular result, are measuring subjective variables, and have an incentive to produce data that confirm predictions. To minimize bias, it is good practice to work "blind," meaning that experimenters are unaware of the identity or treatment group of their subjects while conducting research. Here, using text mining and a literature review, we find evidence that blind protocols are uncommon in the life sciences and that nonblind studies tend to report higher effect sizes and more significant p-values. We discuss methods to minimize bias and urge researchers, editors, and peer reviewers to keep blind protocols in mind.

| S-EPMC4496034 | biostudies-literature

OmicsDI is part of the ELIXIR infrastructure

OmicsDI is an Elixir interoperability service. Learn more ›

Tweets

OmicsDI Databases

PRIDE
PeptideAtlas
MassIVE
JPOST Repository
Physiome Model Repository

EGA
EVA
ENA
LINCS
PAXDB
Cell Collective

MetaboLights
Metabolomics Workbench
MetabolomeExpress
GNPS
BioModels
FAIRDOMHub

ArrayExpress
dbGaP
ExpressionAtlas
GEO
NODE

Information

Databases
Help
API
Contact us
Code on GitHub
Terms of use
Submit Data