UNi Logo
UNi Logo UNi Logo UNi Logo UNi Logo

Über diesen Blog.

Hier schreiben Wissenschaftler*innen der Universität Oldenburg und Gastautor*innen darüber, wie sich Gesellschaften selbst wahrnehmen und thematisieren, sich ihrer jeweiligen Gegenwart vergewissern und dabei in die Zukunft entwerfen.

Wie stehen diese Selbstwahrnehmungen und -entwürfe mit Institutionen, Medien und Techniken zur Gestaltung von Natur, Gesellschaft und Subjektivität in Verbindung? Wie modellieren sie den lebensweltlichen Alltag und halten Menschen zu einem bestimmten Verhalten an? Wie werden diese Interventionen in das Gegebene begründet und legitimiert, aber auch kritisiert, verworfen oder unterlaufen?

Diesen Fragen, deren interdisziplinäre Reflexion eines der zentralen Anliegen des Wissenschaftlichen Zentrums „Genealogie der Gegenwart“ ist, gehen die Blogger aus unterschiedlichen Fachperspektiven und Tätigkeitszusammenhängen mit Blick auf kontrovers verhandelte Themen wie Migration, Ungleichheit, Digitalisierung, Kriminalität, Gesundheit und Ökologie nach.

05.08.2022
Chatstep visitors

Conscious of this new unbalanced proportion out-of female and male examples into the all of our data, i after that investigated anticipate results round the intercourse

von Team

Shade portray anticipate on the CpG websites contained in this specific genomic regions since denoted regarding legend. Each accuracy–bear in mind contour means an average precision–bear in mind to have forecast to your held-away set per of one’s ten repeated random subsamples. (D) Two-dimensional histogram of predicted methylation membership in the place of fresh methylation account. x- and you can y-axes portray assayed as opposed to predicted ? viewpoints, correspondingly. Shade portray the density of each and every matrix unit, averaged over all forecasts having one hundred individuals. CGI, CpG isle; Gene_pos, genomic position; k-NN, k-nearest locals classifier; ROC, person performing trait; seq_assets, series features; SVM, assistance vector servers; TFBS, transcription factor binding web site; HM, histone amendment marks; ChromHMM, chromatin states, given that outlined of the ChromHMM app .

Cross-test prediction

To determine just how predictive methylation pages was round the trials, we quantified the fresh new generalization error of our classifier genome-wide round the people. Specifically, i trained all of our classifier with the 10,000 sites in one personal, and predicted methylation standing for everybody CpG internet sites towards the almost every other 99 some one. The fresh classifier’s show try extremely uniform all over anyone (Extra file step 1: Shape S4), recommending see your face-specific covariates – different size of cell types, including – don’t limitation prediction reliability. This new classifier’s overall performance is highly uniform when studies towards ladies and you will forecasting CpG web site methylation updates in boys, and you will the other way around (Extra file step 1: Contour S5).

To test new awareness in our classifier toward amount of CpG sites regarding the knowledge place, we investigated brand new forecast show for different studies set systems. We learned that studies set that have more than step 1,one hundred thousand CpG internet sites had fairly equivalent performance (Additional document step 1: Contour S6). Throughout these experiments, i utilized an exercise set measurements of 10,100000, so you’re able to struck a balance ranging from adequate quantities of degree examples and you can computational tractability.

Cross-platform prediction

So you’re able to assess group round the system and you will phone-kind of heterogeneity, we investigated the brand new classifier’s overall performance to the WGBS research [59,60]. Particularly, we classified per CpG site in a good WGBS try based on whether you to definitely CpG webpages are assayed towards the 450K array (450K site) or perhaps not (non 450K web site); nearby internet on the WGBS research are websites which might be adjoining toward genome whenever both are 450K internet sites. I have fun with that WGBS try out of b-cells, that fits specific proportion each and every entire bloodstream attempt; i remember that the brand new 450K number entire bloodstream samples often have heterogeneous phone versions compared to this new WGBS investigation. Complete, we come across a much higher ratio of hypomethylated CpG internet sites on the fresh new 450K selection in line with the fresh WGBS studies (Most document step one: Contour S7) of the disproportionate image of hypomethylated CpG web sites in this CGIs chatstep towards 450K variety.

First, we investigated cross-platform prediction, training our classifier on a 450K array sample and testing on WGBS data. We trained the classifier on 10,000 CpG sites in the 450K array samples, and then we tested on 100,000 CpG sites in WGBS data twice – once restricting the test set to 450K sites and once restricting the test set to non 450K sites. We repeated this experiment ten times. Next, we performed the same experiment but trained and tested on the WGBS data. Because the proportion of hypomethylated and hypermethylated sites was imbalanced for CpG sites not on the 450K array, we used a precision–recall curve instead of a ROC curve to measure the prediction performance . We used all 122 features and considered prediction of inverse CpG status \(<\hat>> = -(\tau – 1)\) in this experiment, to assess the quality of the predictions for the less frequent class of hypomethylated CpG sites.

Diskussion einblenden/ausblenden

Bis jetzt noch keine Kommentare.

Einen Kommentar abgeben