Tumor and reproductive traits are linked by RNA metabolism genes in the mouse ovary: a transcriptome-phenotype association analysis

Abstract

BackgroundThe link between reproductive life history and incidence of ovarian tumors is well known. Periods of reduced ovulations may confer protection against ovarian cancer. Using phenotypic data available for mouse, a possible association between the ovarian transcriptome, reproductive records and spontaneous ovarian tumor rates was investigated in four mouse inbred strains. NIA15k-DNA microarrays were employed to obtain expression profiles of BalbC, C57BL6, FVB and SWR adult ovaries.ResultsLinear regression analysis with multiple-test control (adjusted p ≤ 0.05) resulted in ovarian tumor frequency (OTF) and number of litters (NL) as the top-correlated among five tested phenotypes. Moreover, nearly one-hundred genes were coincident between these two traits and were decomposed in 76 OTF(–) NL(+) and 20 OTF(+) NL(–) genes, where the plus/minus signs indicate the direction of correlation. Enriched functional categories were RNA-binding/mRNA-processing and protein folding in the OTF(–) NL(+) and the OTF(+) NL(–) subsets, respectively. In contrast, no associations were detected between OTF and litter size (LS), the latter a measure of ovulation events in a single estrous cycle.ConclusionLiterature text-mining pointed to post-transcriptional control of ovarian processes including oocyte maturation, folliculogenesis and angiogenesis as possible causal relationships of observed tumor and reproductive phenotypes. We speculate that repetitive cycling instead of repetitive ovulations represent the actual link between ovarian tumorigenesis and reproductive records.

Dysregulation of mitotic machinery genes precedes genome instability during spontaneous pre-malignant transformation of mouse ovarian surface epithelial cells

“…Regarding down-regulated genes at p14, the top enriched function was mRNA processing covering 11 genes (Table 1), 8 of which were correlated to number of litters and to ovarian tumor frequency in a previous transcriptomic study of our laboratory aimed to associate reproductive parameters and spontaneous tumor rates across 4 mice strains [39]. RNA processing has been increasingly connected to the DNA damage response [40], which in our results links to apoptosis through Nupr1 and Cdip1 (Table 1), the latter a regulator of TNF-alpha-mediated, p53-dependent apoptosis [41].…”

“…Paraspeckles are emerging as key regulators of gene expression at the post-transcriptional level by its ability to sequestrate certain proteins [44] and retain mature mRNAs in the nucleus [45, 46]. In addition, Malat1 and Neat1 were negatively correlated to ovarian tumor frequency , i.e., their levels were minimal in the ovaries of mouse strains displaying the highest spontaneous tumor rates [39]. Downregulation of Malat1 in this MOSE model along with its inverse correlation with spontaneous ovarian tumors in mice, is consistent with a recent report suggesting a tumor-suppressor role of Malat1 in gliomas [31].…”

Applying Citizen Science to Gene, Drug, Disease Relationship Extraction from Biomedical Abstracts

“…Making that knowledge more accessible and computable could save researchers time, effort, and resources (Yang et al, 2016) (Zhu et al, 2013). Researchers have effectively mined slices of the biomedical literature to identify potential treatments for Raynaud's syndrome (Swanson 1986), drug candidates for Alzheimer's disease (Li, Zhu and Chen, 2009), and potential mechanisms of ovarian oncogenesis (Urzúa et al, 2010). Given the large potential to make valuable inferences and the large volume of literature, many researchers have turned to information extraction algorithms to harvest information in biomedical texts and improve the value of existing data resources (Murray-Rust, 2017) (Pletscher-Frankild et al, 2015).…”

Automatic design of decision-tree induction algorithms tailored to flexible-receptor docking data

“…We name this type of receptor representation fully flexible-receptor (FFR) model [6,7], and we investigate this methodology with target receptor InhA enzyme from Mycobacterium tuberculosis [8] (Mtb), which was modeled as a set of 3,100 snapshots derived from a 3.1 ns MD simulation trajectory [9]. For that, we generated molecular docking data sets with data from docking simulations of FFR-InhA [10] to six different ligands: nicotinamide adenine dinucleotide ( NADH ) [8], triclosan ( TCL ) [11], pentacyano(isoniazid)ferrate(II) ( PIF ) [12], ethionamide ( ETH ) [13], Isoniazid ( INH ) [14], and Triclosan derivative 20 ( JPM ) [15].…”

“…With the resulting induced models from these automatically-designed algorithms, we expect that the inferred knowledge will help us to point out which of the conformations that were generated by the fully flexible-receptor model are more promising to future docking experiments. This, in turn, allows a reduction of the flexible-receptor model dimensionality and permit faster docking simulations of flexible receptors [6]. We analyze whether the decision trees generated by the automatically-designed algorithms have higher predictive accuracy and are more comprehensible than decision trees generated by state-of-the-art decision-tree induction algorithm, C4.5 [20].…”