Inferring feature importance with uncertainties with application to large genotype data.

Estimating feature importance, which is the contribution of a prediction or several predictions due to a feature, is an essential aspect of explaining data-based models. Besides explaining the model itself, an equally relevant question is which features are important in the underlying data generating process. We present a Shapley-value-based framework for inferring the importance of individual features, including uncertainty in the estimator. We build upon the recently published model-agnostic feature importance score of SAGE (Shapley additive global importance) and introduce Sub-SAGE. For tree-based models, it has the advantage that it can be estimated without computationally expensive resampling. We argue that for all model types the uncertainties in our Sub-SAGE estimator can be estimated using bootstrapping and demonstrate the approach for tree ensemble methods. The framework is exemplified on synthetic data as well as large genotype data for predicting feature importance with respect to obesity.

Exploring shared genetics between maximal oxygen uptake and disease: the HUNT study.

Low cardiorespiratory fitness, measured as maximal oxygen uptake (V̇o2max), is associated with all-cause mortality and disease-specific morbidity and mortality and is estimated to have a large genetic component (∼60%). However, the underlying mechanisms explaining the associations are not known, and no association study has assessed shared genetics between directly measured V̇o2max and disease. We believe that identifying the mechanisms explaining how low V̇o2max is related to increased disease risk can contribute to prevention and therapy. We used a phenome-wide association study approach to test for shared genetics. A total of 64,479 participants from the Trøndelag Health Study (HUNT) were included. Genetic variants previously linked to V̇o2max were tested for association with diseases related to the cardiovascular system, diabetes, dementia, mental disorders, and cancer as well as clinical measurements and biomarkers from HUNT. In the total population, three single-nucleotide polymorphisms (SNPs) in and near the follicle-stimulating hormone receptor gene (FSHR) were found to be associated (false discovery rate < 0.05) with serum creatinine levels and one intronic SNP in the Rap-associating DIL domain gene (RADIL) with diabetes type 1 with neurological manifestations. In males, four intronic SNPs in the PBX/knotted homeobox 2 gene (PKNOX2) were found to be associated with endocarditis. None of the association tests in the female population reached overall statistical significance; the associations with the lowest P values included other cardiac conduction disorders, subdural hemorrhage, and myocarditis. The results might suggest shared genetics between V̇o2max and disease. However, further effort should be put into investigating the potential shared genetics between inborn V̇o2max and disease in larger cohorts to increase statistical power.NEW & NOTEWORTHY To our knowledge, this is the first genetic association study exploring how genes linked to cardiorespiratory fitness (CRF) relate to disease risk. By investigating shared genetics, we found indications that genetic variants linked to directly measured CRF also affect the level of blood creatinine, risk of diabetes, and endocarditis. Less certain findings showed that genetic variants of high CRF might cause lower body mass index, healthier HDL cholesterol, and lower resting heart rate.

Saddlepoint approximations to score test statistics in logistic regression for analyzing genome-wide association studies.

We investigate saddlepoint approximations of tail probabilities of the score test statistic in logistic regression for genome-wide association studies. The inaccuracy in the normal approximation of the score test statistic increases with increasing imbalance in the response and with decreasing minor allele counts. Applying saddlepoint approximation methods greatly improve the accuracy, even far out in the tails of the distribution. By using exact results for a simple logistic regression model, as well as simulations for models with nuisance parameters, we compare double saddlepoint methods for computing two-sided P $$ P $$ -values and mid- P $$ P $$ -values. These methods are also compared to a recent single saddlepoint procedure. We investigate the methods further on data from UK Biobank with skin and soft tissue infections as phenotype, using both common and rare variants.

Low serum lipocalin-2 in pregnant women with systemic lupus erythematosus.

OBJECTIVES: Systemic lupus erythematosus (SLE) pregnancies are considered high-risk due to risk of disease flare and pregnancy complications. A more in-depth understanding of the immunological alterations in SLE patients during pregnancy and identification of predictive biomarkers may help to achieve stable disease and to avoid pregnancy complications. Lipocalin-2 (LCN2) has been implicated as a potential biomarker for rheumatic diseases and preeclampsia, but remains unexplored in SLE pregnancies. METHODS: We measured LCN2 levels in serum samples from SLE pregnancies (n=25) at seven different time points. Samples were taken preconception, in each trimester, at 6 weeks, 6 months and 12 months postpartum. Serum LCN2 levels were compared to samples from rheumatoid arthritis (RA) (n=27) and healthy (n=18) pregnancies at each time point using t-test, and for all time points using a linear mixed effects model. In addition, we investigated the association between LCN2 levels and disease activity, CRP, kidney function, BMI, treatment regimen and adverse pregnancy outcome for SLE and RA patients. RESULTS: We found significantly lower serum LCN2 levels throughout pregnancy in SLE patients with quiescent disease compared to RA and healthy pregnancies. We did not find an association between serum LCN2 and disease activity or adverse pregnancy outcome in SLE pregnancies. CONCLUSIONS: In a population of SLE women with low disease activity we have not found evidence that serum LCN2 levels predict disease activity or adverse pregnancy outcomes. Further studies are needed to elucidate a possible biological role of low LCN2 levels in SLE pregnancies.

A new method for exploring gene-gene and gene-environment interactions in GWAS with tree ensemble methods and SHAP values.

BACKGROUND: The identification of gene-gene and gene-environment interactions in genome-wide association studies is challenging due to the unknown nature of the interactions and the overwhelmingly large number of possible combinations. Parametric regression models are suitable to look for prespecified interactions. Nonparametric models such as tree ensemble models, with the ability to detect any unspecified interaction, have previously been difficult to interpret. However, with the development of methods for model explainability, it is now possible to interpret tree ensemble models efficiently and with a strong theoretical basis. RESULTS: We propose a tree ensemble- and SHAP-based method for identifying as well as interpreting potential gene-gene and gene-environment interactions on large-scale biobank data. A set of independent cross-validation runs are used to implicitly investigate the whole genome. We apply and evaluate the method using data from the UK Biobank with obesity as the phenotype. The results are in line with previous research on obesity as we identify top SNPs previously associated with obesity. We further demonstrate how to interpret and visualize interaction candidates. CONCLUSIONS: The new method identifies interaction candidates otherwise not detected with parametric regression models. However, further research is needed to evaluate the uncertainties of these candidates. The method can be applied to large-scale biobanks with high-dimensional data.

Mobile Sensing in Substance Use Research: A Scoping Review.

Background:Addictive disorders and substance use are significant health challenges worldwide, and relapse is a core component of addictive disorders. The dynamics surrounding relapse and especially the immediate period before it occurs is only partly understood, much due to difficulties collecting reliable and sufficient data from this narrow period. Mobile sensing has been an important way to improve data quality and enhance predictive capabilities for symptom worsening within physical and mental health care, but is less developed within substance use research.Methodology:This scoping review aimed to reviewing the currently available research on mobile sensing of substance use and relapse in substance use disorders. The search was conducted in January 2019 using PubMed and Web of Science.Results:Six articles were identified, all concerning subjects using alcohol. In the studies a range of mobile sensors and derived aggregated features were employed. Data collected through mobile sensing were predominantly used to make dichotomous inference on ongoing substance use or not and in some cases on the quantity of substance intake. Only one of the identified studies predicted later substance use. A range of statistical machine learning techniques was employed.Conclusions:The research on mobile sensing in this field remains scarce. The issues requiring further attention include more research on clinical populations in naturalistic settings, use of a priori knowledge in statistical modeling, focus on prediction of substance use rather than purely identification, and finally research on other substances than alcohol.

Powerful extreme phenotype sampling designs and score tests for genetic association studies.

We consider cross-sectional genetic association studies (common and rare variants) where non-genetic information is available or feasible to obtain for N individuals, but where it is infeasible to genotype all N individuals. We consider continuously measurable Gaussian traits (phenotypes). Genotyping n < N extreme phenotype individuals can yield better power to detect phenotype-genotype associations, as compared to randomly selecting n individuals. We define a person as having an extreme phenotype if the observed phenotype is above a specified threshold or below a specified threshold. We consider a model where these thresholds can be tailored to each individual. The classical extreme sampling design is to set equal thresholds for all individuals. We introduce a design (z-extreme sampling) where personalized thresholds are defined based on the residuals of a regression model including only non-genetic (fully available) information. We derive score tests for the situation where only n extremes are analyzed (complete case analysis) and for the situation where the non-genetic information on N - n non-extremes is included in the analysis (all case analysis). For the classical design, all case analysis is generally more powerful than complete case analysis. For the z-extreme sample, we show that all case and complete case tests are equally powerful. Simulations and data analysis also show that z-extreme sampling is at least as powerful as the classical extreme sampling design and the classical design is shown to be at times less powerful than random sampling. The method of dichotomizing extreme phenotypes is also discussed.

Clusters. / Klynger.

Not all data sets have explanatory variables and outcomes. The data may nevertheless contain associations that are worth revealing.

What proportion of published research findings are false? / Hvor stor andel av publiserte forskningsresultat er feil?

Not all published research findings are reproducible ­ some because the findings are incorrect. What is the extent of the problem? A number of researchers have attempted to estimate how often published findings are false. They have used widely different approaches.

Plant defence responses in oilseed rape MINELESS plants after attack by the cabbage moth Mamestra brassicae.

The Brassicaceae family is characterized by a unique defence mechanism known as the 'glucosinolate-myrosinase' system. When insect herbivores attack plant tissues, glucosinolates are hydrolysed by the enzyme myrosinase (EC 3.2.1.147) into a variety of degradation products, which can deter further herbivory. This process has been described as 'the mustard oil bomb'. Additionally, insect damage induces the production of glucosinolates, myrosinase, and other defences. Brassica napus seeds have been genetically modified to remove myrosinase-containing myrosin cells. These plants are termed MINELESS because they lack myrosin cells, the so-called toxic mustard oil mines. Here, we examined the interaction between B. napus wild-type and MINELESS plants and the larvae of the cabbage moth Mamestra brassicae. No-choice feeding experiments showed that M. brassicae larvae gained less weight and showed stunted growth when feeding on MINELESS plants compared to feeding on wild-type plants. M. brassicae feeding didn't affect myrosinase activity in MINELESS plants, but did reduce it in wild-type seedlings. M. brassicae feeding increased the levels of indol-3-yl-methyl, 1-methoxy-indol-3-yl-methyl, and total glucosinolates in both wild-type and MINELESS seedlings. M. brassicae feeding affected the levels of glucosinolate hydrolysis products in both wild-type and MINELESS plants. Transcriptional analysis showed that 494 and 159 genes were differentially regulated after M. brassicae feeding on wild-type and MINELESS seedlings, respectively. Taken together, the outcomes are very interesting in terms of analysing the role of myrosin cells and the glucosinolate-myrosinase defence system in response to a generalist cabbage moth, suggesting that similar studies with other generalist or specialist insect herbivores, including above- and below-ground herbivores, would be useful.

Robust methods to detect disease-genotype association in genetic association studies: calculate p-values using exact conditional enumeration instead of simulated permutations or asymptotic approximations.

In genetic association studies, detecting disease-genotype association is a primary goal. We study seven robust test statistics for such association when the underlying genetic model is unknown, for data on disease status (case or control) and genotype (three genotypes of a biallelic genetic marker). In such studies, p-values have predominantly been calculated by asymptotic approximations or by simulated permutations. We consider an exact method, conditional enumeration. When the number of simulated permutations tends to infinity, the permutation p-value approaches the conditional enumeration p-value, but calculating the latter is much more efficient than performing simulated permutations. We have studied case-control sample sizes with 500-5000 cases and 500-15,000 controls, and significance levels from 5 × 10(-8) to 0.05, thus our results are applicable to genetic association studies with only a few genetic markers under study, intermediate follow-up studies, and genome-wide association studies. Our main findings are: (i) If all monotone genetic models are of interest, the best performance in the situations under study is achieved for the robust test statistics based on the maximum over a range of Cochran-Armitage trend tests with different scores and for the constrained likelihood ratio test. (ii) For significance levels below 0.05, for the test statistics under study, asymptotic approximations may give a test size up to 20 times the nominal level, and should therefore be used with caution. (iii) Calculating p-values based on exact conditional enumeration is a powerful, valid and computationally feasible approach, and we advocate its use in genetic association studies.

The duration of gastrin treatment affects global gene expression and molecular responses involved in ER stress and anti-apoptosis.

BACKGROUND: How cells decipher the duration of an external signal into different transcriptional outcomes is poorly understood. The hormone gastrin can promote a variety of cellular responses including proliferation, differentiation, migration and anti-apoptosis. While gastrin in normal concentrations has important physiological functions in the gastrointestine, prolonged high levels of gastrin (hypergastrinemia) is related to pathophysiological processes. RESULTS: We have used genome-wide microarray time series analysis and molecular studies to identify genes that are affected by the duration of gastrin treatment in adenocarcinoma cells. Among 403 genes differentially regulated in transiently (gastrin removed after 1 h) versus sustained (gastrin present for 14 h) treated cells, 259 genes upregulated by sustained gastrin treatment compared to untreated controls were expressed at lower levels in the transient mode. The difference was subtle for early genes like Junb and c-Fos, but substantial for delayed and late genes. Inhibition of protein synthesis by cycloheximide was used to distinguish between primary and secondary gastrin regulated genes. The majority of gastrin upregulated genes lower expressed in transiently treated cells were primary genes induced independently of de novo protein synthesis. This indicates that the duration effect of gastrin treatment is mainly mediated via post-translational signalling events, while a smaller fraction of the differentially expressed genes are regulated downstream of primary transcriptional events. Indeed, sustained gastrin treatment specifically induced prolonged ERK1/2 activation and elevated levels of the AP-1 subunit protein JUNB. Enrichment analyses of the differentially expressed genes suggested that endoplasmic reticulum (ER) stress and survival is affected by the duration of gastrin treatment. Sustained treatment exerted an anti-apoptotic effect on serum starvation-induced apoptosis via a PKC-dependent mechanism. In accordance with this, only sustained treatment induced anti-apoptotic genes like Clu, Selm and Mcl1, while the pro-apoptotic gene Casp2 was more highly expressed in transiently treated cells. Knockdown studies showed that JUNB is involved in sustained gastrin induced expression of the UPR/ER stress related genes Atf4, Herpud1 and Chac1. CONCLUSION: The duration of gastrin treatment affects both intracellular signalling mechanisms and gene expression, and ERK1/2 and AP-1 seem to play a role in converting different durations of gastrin treatment into distinct cellular responses.

Variability in UDP-glucuronosyltransferase genes and morphine metabolism: observations from a cross-sectional multicenter study in advanced cancer patients with pain.

OBJECTIVE: The objective of the present study was to determine whether genetic variability in UDP-glucuronosyltransferase (UGT) genes, together with clinical factors, contribute to variability in morphine glucuronide (M6G and M3G) to morphine serum concentration ratios in patients with advanced cancer receiving chronic morphine therapy. MATERIALS AND METHODS: A total of 41 polymorphisms and predicted haplotypes in the UGT2B7, UGT1A1, and UGT1A8 genes were analyzed in 759 patients who were recruited from the European Pharmacogenetic Opioid Study and received chronic morphine therapy by the oral route (n=635) or parenterally (n=124). The administration groups were analyzed separately by multiple linear regression analyses. RESULTS: Two haplotypes in UGT1A1/UGT1A8 were weak predictors of reduced M6G/morphine and M3G/morphine serum ratios after oral administration (false discovery rate-corrected P-values<0.1). No effect of genotype was seen in the parenteral group. Of the clinical variables (age, sex, BMI, renal function, Karnofsky performance status, and presence of liver metastases), renal function was the major contributor to variation in serum concentration ratios. Concomitant administration of paracetamol predicted significantly higher morphine metabolic ratios after oral administration of morphine (false discovery rate-corrected P-values<2.1E-12). The regression models explained about 35% of the total variability in the data. CONCLUSION: Genetic variation in the UGT genes together with clinical factors influence morphine metabolic ratios in patients with advanced cancer disease and who are scheduled with oral morphine. This information may be included in future research that develop and test new classification systems for opioid treatment in patients with advanced cancer.

Exploring Dietary Intake in Adults with Severe Obesity and Associations with the FTO rs9939609 Genotypes.

Background: Few have studied the associations between rs9939609 genotypes in the obesity candidate locus FTO and energy and nutrient intakes and meal frequencies in adults with severe obesity. We are unaware of studies that have assessed adherence to key dietary recommendations in this population, at least in Norway. Increased knowledge of genotype associations with dietary factors could improve personalized obesity therapy. Objectives: The present study aimed to explore how the rs9939609 genotypes associate with dietary variables and adherence to key dietary recommendations in a sample of adults with severe obesity. Methods: A cross-sectional observation study designed to have similar numbers of participants with genotypes TT, AT, and AA included 100 patients (70% women) with median (25th, 75th percentile) age 42 (32, 50) y and BMI 42.8 (39.5, 46.4) kg/m2. We assessed intakes of food groups, energy, and macro- and micronutrients from three 24-h dietary recalls and meal frequencies. Genotype associations were analyzed using regression analyses. Reported intakes were evaluated against national diet recommendations. Results: Using a significance level of 0.01, we found no genotype associations with energy intake, energy density, adherence to recommendations, or meal frequency but tendencies of associations with energy adjusted protein intake (AA > AT, P = 0.037; AT > TT, P = 0.064), food groups milk and cream (AT > TT, P = 0.029), and Mixed dishes (AA > TT, P = 0.039). Few participants complied with recommendations for intakes of whole grains (21%), fruits and vegetables (11%), and fish (37%); however, 67% followed the recommendation to limit added sugar. Less than 20% had recommended intakes of vitamin D and folate. Conclusions: In our patients with severe obesity, we found tendencies of associations between the FTO rs9939609 genotypes and diet but no significant associations at the 0.01 level and below. Few met key food-based diet recommendations, suggesting that the food habits in this population pose an increased risk of nutrient deficiencies. Curr Dev Nutr 2023;xx:xx.

Mobile Assessments of Mood, Cognition, Smartphone-Based Sensor Activity, and Variability in Craving and Substance Use in Patients With Substance Use Disorders in Norway: Prospective Observational Feasibility Study.

BACKGROUND: Patients with substance use disorders (SUDs) are at increased risk for symptom deterioration following treatment, with up to 60% resuming substance use within the first year posttreatment. Substance use craving together with cognitive and mental health variables play important roles in the understanding of the trajectories from abstinence to substance use. OBJECTIVE: This prospective observational feasibility study aims to improve our understanding of specific profiles of variables explaining SUD symptom deterioration, in particular, how individual variability in mental health, cognitive functioning, and smartphone use is associated with craving and substance use in a young adult clinical population. METHODS: In this pilot study, 26 patients with SUDs were included at about 2 weeks prior to discharge from inpatient SUD treatment from 3 different treatment facilities in Norway. Patients underwent baseline neuropsychological and mental health assessments; they were equipped with smartwatches and they downloaded an app for mobile sensor data collection in their smartphones. Every 2 days for up to 8 weeks, the patients were administered mobile ecological momentary assessments (EMAs) to evaluate substance use, craving, mental health, cognition, and a mobile Go/NoGo performance task. Repeated EMAs as well as the smartphone's battery use data were averaged across all days per individual and used as candidate input variables together with the baseline measures in models of craving intensity and the occurrence of any substance use episodes. RESULTS: A total of 455 momentary assessments were completed out of a potential maximum of 728 assessments. Using EMA and baseline data as candidate input variables and craving and substance use as responses, model selection identified mean craving intensity as the most important predictor of having one or more substance use episodes and with variabilities in self-reported impulsivity, mental health, and battery use as significant explanatory variables of craving intensity. CONCLUSIONS: This prospective observational feasibility study adds novelty by collecting high-intensity data for a considerable period of time, including mental health data, mobile cognitive assessments, and mobile sensor data. Our study also contributes to our knowledge about a clinical population with the most severe SUD presentations in a vulnerable period during and after discharge from inpatient treatment. We confirmed the importance of variability in cognitive function and mood in explaining variability in craving and that smartphone usage may possibly add to this understanding. Further, we found that craving intensity is an important explanatory variable in understanding substance use episodes.

Predicting Relapse in Substance Use: Prospective Modeling Based on Intensive Longitudinal Data on Mental Health, Cognition, and Craving.

Patients with severe substance use disorders are often characterized by neurocognitive impairments and elevated mental health symptom load, typically associated with craving intensity and substance use relapse. There is a need to improve the predictive capabilities of when relapse occurs in order to improve substance use treatment. The current paper contains data from 19 patients (seven females) in a long-term inpatient substance use treatment setting over the course of several weeks, with up to three weekly data collections. We collected data from 252 sessions, ranging from 1 to 24 sessions per subject. The subjects reported craving, self-control, and mental health on each occasion. Before starting the repeated data collection, a baseline neuropsychological screening was performed. In this repeated-measures prospective study, the mixed-effects models with time-lagged predictors support a model of substance use craving and relapse being predicted by the baseline reaction time as well as the temporal changes and variability in mental health symptom load, self-control, and craving intensity with moderate to high effect sizes. This knowledge may contribute to more personalized risk assessments and treatments for this group of patients.

A Bayesian hierarchical model for quantitative real-time PCR data.

We present a Bayesian hierarchical model for quantitative real-time polymerase chain reaction (PCR) data, aiming at relative quantification of DNA copy number in different biological samples. The model is specified in terms of a hidden Markov model for fluorescence intensities measured at successive cycles of the polymerase chain reaction. The efficiency of the reaction is assumed to depend on the abundance of the target DNA through fluorescence intensities, and the relationship is specified based on the kinetics of the reaction. The model incorporates the intrinsic random nature of the process as well as measurement error. Taking a Bayesian inferential approach, marginal posterior distributions of the quantities of interest are estimated using Markov chain Monte Carlo. The method is applied to simulated data and an experimental data set.

ADHD and Mental Health Symptoms in the Identification of Young Adults with Increased Risk of Alcohol Dependency in the General Population-The HUNT4 Population Study.

Symptoms of ADHD are strongly associated with alcohol use disorders, and mental health symptoms attenuate this relationship. There is limited knowledge about how specific symptoms of inattentiveness and hyperactivity/impulsivity can explain this association. We aimed to identify self-reported executive cognitive functioning and mental health and variables that may help identify subjects with an elevated risk of alcohol dependence in the general population. Data included 3917 subjects between 19 and 30 years old in the 4th Trøndelag Health Study. The Adult ADHD Self report Scale-Screener, the Hospital Anxiety and Depression Scale, and demographic variables were used as input variables. The alcohol screening instrument CAGE was used as the response variable for binary alcohol dependence risk. We used logistic regression and automated model selection to arrive at our final model that identified sex, age, inattentiveness, hyperactivity/impulsivity symptoms, and anxiety as predictors of having a CAGE score ≥2, achieving an area under the receiver operating characteristic curve of 0.692. A balanced accuracy approach indicated an optimal cut-off of 0.153 with sensitivity 0.55 and specificity 0.74. Despite attrition in the data, our findings may be important in the assessment of individual risk for alcohol dependency and when developing algorithms for risk triage in public health.

The fat mass and obesity-associated (FTO) gene allele rs9939609 and glucose tolerance, hepatic and total insulin sensitivity, in adults with obesity.

The objective of the study was to assess associations of the rs9939609 FTO allele to glucose tolerance, hepatic and total insulin sensitivity (IS) in individuals with obesity. From a low-dose hyperinsulinemic euglycemic clamp with glucose-tracer, hepatic IS was assessed by rates of basal and suppressed glucose appearance (Ra), a measure of endogenous glucose production (EGP), and the hepatic insulin resistance index (HIR). Total IS was assessed by rates of glucose infusion (GIR), disappearance (Rd), and metabolic clearance (MCR). From a meal test we assessed IS by the Matsuda index and glucose tolerance by glucose and insulin measurements in the fasted state and postprandially for 2.5 h. The meal test was performed in 97 healthy individuals with BMI ≥35 in similar-sized risk-allele groups (n = 32 T/T, 31 A/T, and 34 A/A), and 79 of them performed the clamp. We analyzed outcomes separately for males and females, and adjusted glucose Ra, Rd, MCR, GIR, and HIR for fat mass. We did not find genotype effects on EGP. Among males, genotype A/A was associated with a significantly lower glucose Rd, MCR, and Matsuda index score relative to genotype T/T. Glucose tolerance was significantly lower in males with genotype A/T vs. T/T and A/A. For females, there were no genotype effects on hepatic or total IS, or on glucose tolerance. Independently of genotypes, females displayed a significantly better hepatic and total IS, and better glucose tolerance than males. We conclude that in subjects with similar obesity we did not register any FTO risk-allele effect on hepatic IS. A FTO risk-allele effect on total IS was registered in males only, findings which need to be reproduced in further studies. Results confirm marked differences in IS between the biological sexes and extend present knowledge by demonstrating a lower endogenous glucose production in females vs. males in uniformly obese individuals.

Functional studies on transfected cell microarray analysed by linear regression modelling.

Transfected cell microarray is a promising method for accelerating the functional exploration of the genome, giving information about protein function in the living cell. The microarrays consist of clusters of cells (spots) overexpressing or silencing a particular gene product. The subsequent analysis of the phenotypic consequences of such perturbations can then be detected using cell-based assays. The focus in the present study was to establish an experimental design and a robust analysis approach for fluorescence intensity data, and to address the use of replicates for studying regulation of gene expression with varying complexity and effect size. Our analysis pipeline includes measurement of fluorescence intensities, normalization strategies using negative control spots and internal control plasmids, and linear regression (ANOVA) modelling for estimating biological effects and calculating P-values for comparisons of interests. Our results show the potential of transfected cell microarrays in studying complex regulation of gene expression by enabling measurement of biological responses in cells with overexpression and downregulation of specific gene products, combined with the possibility of assaying the effects of external stimuli. Simulation experiments show that transfected cell microarrays can be used to reliably detect even quantitatively minor biological effects by including several technical and experimental replicates.