The influence of missing value imputation on detection of differentially expressed genes from microarray data

MOTIVATION: Missing values are problematic for the analysis of microarray data. Imputation methods have been compared in terms of the similarity between imputed and true values in simulation experiments and not of their influence on the final analysis. The focus has been on missing at random, while entries are missing also not at random. RESULTS: We investigate the influence of imputation on the detection of differentially expressed genes from cDNA microarray data. We apply ANOVA for microarrays and SAM and look to the differentially expressed genes that are lost because of imputation. We show that this new measure provides useful information that the traditional root mean squared error cannot capture. We also show that the type of missingness matters: imputing 5% missing not at random has the same effect as imputing 10-30% missing at random. We propose a new method for imputation (LinImp), fitting a simple linear model for each channel separately, and compare it with the widely used KNNimpute method. For 10% missing at random, KNNimpute leads to twice as many lost differentially expressed genes as LinImp. AVAILABILITY: The R package for LinImp is available at http://folk.uio.no/idasch/imp.     

Kir6.2 mutations causing neonatal diabetes provide new insights into Kir6.2-SUR1 interactions

ATP-sensitive K(+) (K(ATP)) channels, comprised of pore-forming Kir6.2 and regulatory SUR1 subunits, play a critical role in regulating insulin secretion. Binding of ATP to Kir6.2 inhibits, whereas interaction of MgATP with SUR1 activates, K(ATP) channels. We tested the functional effects of two Kir6.2 mutations (Y330C, F333I) that cause permanent neonatal diabetes mellitus, by heterologous expression in Xenopus oocytes. Both mutations reduced ATP inhibition and increased whole-cell currents, which in pancreatic beta-cells is expected to reduce insulin secretion and precipitate diabetes. The Y330C mutation reduced ATP inhibition both directly, by impairing ATP binding (and/or transduction), and indirectly, by stabilizing the intrinsic open state of the channel. The F333I mutation altered ATP binding/transduction directly. Both mutations also altered Kir6.2/SUR1 interactions, enhancing the stimulatory effect of MgATP (which is mediated via SUR1). This effect was particularly dramatic for the Kir6.2-F333I mutation, and was abolished by SUR1 mutations that prevent MgATP binding/hydrolysis. Further analysis of F333I heterozygous channels indicated that at least three SUR1 must bind/hydrolyse MgATP to open the mutant K(ATP) channel.     

Muscle metabolism and quality (MQI) in prediabetic sedentary man

Twelve pairs of healthy sedentary males matched for their body mass index (BMI) with either a low insulin response (LIR; a stage of prediabetes) or a high (HIR; controls) to a standardized glucose infusion test (GIT) were studied in respect to their exercise capacities (W_OBLA, W_SL and relative W_OBLA: W_OBLA × W_SL-1 × 100), muscle fiber composition (%ST), muscle citrate synthase activity (CS), muscle ubiquinone (MUQ), MUQ over %ST (muscle quality index, MQI), and peripheral insulin sensitivity (PIS) as described with insulin-clamp techniques (SIGITmean). LIR and HIR displayed normal PIS and positive relationships versus exercise capacity. LIR's but not HIR's relative W_OBLA was related to CS as earlier only documented in endurance athletes but at a lower level than both in HIR and athletes. This pointed at a poor peripheral oxygen delivery in LIR. LIR's MQI decreased relative to HIR's the higher the muscle CS indicating radical related muscle trauma in LIR as in athletes. LIR representing prediabetes described muscle anomalies, which could represent prestages of the lesions observed in type-2 diabetes. They are claimed to be responsible for insulin-, glucose-, lipid-resistance, and peripheral circulatory resistance.     

Human pepsinogen A (PGA): an informative gene complex located at 11q13

Summary Human pepsinogen (PGA) exhibits extensive polymorphism that can be detected both at the protein and the DNA level. We describe here two restriction fragment length polymorphisms, EcoRI and BglII, which provide for the detection of three of the most common PGA haplotypes (A, B, and C) in the United States population. The relationship of these polymorphisms to each PGA haplotype was determined by analysis of DNA from individuals exhibiting the corresponding protein phenotypes and by analysis of a series of human × mouse somatic cell hybrids containing the individual chromosome 11 homologous from heterozygous individuals exhibiting the AB and AC protein phenotypes. The use of the BglII polymorphism in combination with previously described EcoRI polymorphism provides a very informative marker of 11q13.     

Inactivation of eucaryotic ribosomes by the toxic plant proteins abrin and ricin

The effect of abrin and ricin on protein synthesizing systems from different sources was studied. The protein synthesis in a cell-free system from rabbit reticulocytes and from rat liver was strongly inhibited by the toxins, whereas a system from E. coli was not affected. Separate treatments of ribosomes and postribosomal supernatant from a rabbit reticulocyte lysate showed that the site of action of the toxins is on the ribosomes. The inactivation of the rabbit reticulocyte lysate by the toxins was a function of the incubation time and temperature. Protein synthesis was not necessary for the toxins to exert their effect. The data indicate that abrin and ricin act only on the eucaryotic type of ribosomes, and that they exert their effect by enzymatic action.     

Mitotic cells form actin-based bridges with adjacent cells to provide intercellular communication during rounding

In order to achieve accurate chromosome segregation, eukaryotic cells undergo a dramatic change in morphology to obtain a spherical shape during mitosis. Interphase cells communicate directly with each other by exchanging ions and small molecules via gap junctions, which have important roles in controlling cell growth and differentiation. As cells round up during mitosis, the gap junctional communication between mitotic cells and adjacent interphase cells ceases. Whether mitotic cells use alternative mechanisms for mediating direct cell-cell communication during rounding is currently unknown. Here, we have studied the mechanisms involved in the remodeling of gap junctions during mitosis. We further demonstrate that mitotic cells are able to form actin-based plasma membrane bridges with adjacent cells during rounding. These structures, termed “mitotic nanotubes,” were found to be involved in mediating the transport of cytoplasm, including Rab11-positive vesicles, between mitotic cells and adjacent cells. Moreover, a subpool of the gap-junction channel protein connexin43 localized in these intercellular bridges during mitosis. Collectively, the data provide new insights into the mechanisms involved in the remodeling of gap junctions during mitosis and identify actin-based plasma membrane bridges as a novel means of communication between mitotic cells and adjacent cells during rounding.     

Preconditioning with estradiol abolishes its neuroprotection in cerebellar neurons

Certain steroids are neuroprotective because they are modulators of neuronal activity or ROS scavengers. We examined neuroprotection following glutamate-induced excitotoxicity in cerebellar granule neuron cultures. 17β-Estradiol, 17α-estradiol (nonestrogenic), or vitamins C+E were equally neuroprotective when coadministered with glutamate, consistent with protection by ROS scavenging. Progesterone protected mainly by an action on GABA-A receptors. Since exogenous antioxidants may influence the level of glutathione, the main endogenous antioxidant in neurons, we investigated if a preconditioning period with the neuroprotectors changed their efficacy as protectors. The neuroprotection by 17β-estradiol and 17α-estradiol, but not progesterone or vitamins C+E, was almost abolished following a preconditioning period of 24 h. This reduction was accompanied by an inhibition of the γ-glutamylcysteine synthetase promoter, and a reduced level of glutathione when preconditioning was combined with the subsequent glutamate exposure. Thus, vitamins C+E and progesterone were more effective long-term neuroprotectors, since preconditioning did not reduce glutathione.