PMC:4979052 / 32005-34289
Annnotations
{"target":"https://pubannotation.org/docs/sourcedb/PMC/sourceid/4979052","sourcedb":"PMC","sourceid":"4979052","source_url":"https://www.ncbi.nlm.nih.gov/pmc/4979052","text":"3.5. Guanine Effects\nIt was previously found that runs of guanines within the probe sequence, particularly runs of guanines as long or longer than three, significantly affect the obtained signal intensities [34]. We showed that this (GGG)1 effect propagates through the various preprocessing methods of microarray data analysis and can lead to biased expression estimates [25]. The origin of the (GGG)1 effect lies in the formation G-quadruplex structures. The formation of duplexes between negatively charged nucleic acids in general, and between G-quadruplexes in particular, depends on the ionic strength and thus on the employed solution buffer [37]. This dependency on the ionic strength also applies to hybridization reactions on solid surfaces [38].\nWe here define the strength of the guanine effect in terms of the intensity increase due to the (GGG)1 motif as follows (6) δI(GGG1)=〈logIp〉ξp1,3=(GGG)−〈logIp〉(TTT)∈ξp⋅ A value of δI(GGG)1 = 0.3 thus reflects an on the average 100.3 ≈ 2 times as large intensity of probes containing the (GGG)1 motif compared to probes containing (TTT) anywhere in their sequence. The average intensity of (TTT) containing probes here serves as appropriate baseline.\nFigure 4b shows the density distribution of the δI(GGG)1 parameter which varies between 0 \u003c δI(GGG)1 \u003c 0.3 for qc-included samples. We consider samples with a threshold of δI(GGG)1 \u003e 0.25 to be significantly affected by GGG1 effect—this reflects an intensity increase of +75% and would lead to a significant bias in the expression estimates of transcripts being targeted by the respective probes. Accordingly, 254 (3.1%) of the samples have a GGG1 related intensity bias. Many of them are removed by strict quality control; only 63 (1.1%) of qc-included samples are above the threshold.\nAssessing the correlation of δI(GGG)1 with the first five principal components of the HumanExpressionAtlas expression data we observe correlation coefficients between 0.14 \u003c |r| \u003c 0.19. Consequently, guanine effects have only a minor impact on the common patterns in the expression space. Note that the RMA method was used for the preprocessing of the HumanExpressionAtlas calibration, and we showed previously that its expression estimates are in general susceptible to GGG1 effects [25].\n","divisions":[{"label":"Title","span":{"begin":0,"end":20}}],"tracks":[{"project":"2_test","denotations":[{"id":"27600351-18718931-69477566","span":{"begin":650,"end":652},"obj":"18718931"},{"id":"27600351-18381819-69477567","span":{"begin":752,"end":754},"obj":"18381819"}],"attributes":[{"subj":"27600351-18718931-69477566","pred":"source","obj":"2_test"},{"subj":"27600351-18381819-69477567","pred":"source","obj":"2_test"}]}],"config":{"attribute types":[{"pred":"source","value type":"selection","values":[{"id":"2_test","color":"#ec9397","default":true}]}]}}