biostudies-arrayexpress00640Brian RingHomo sapienshttps://www.ebi.ac.uk/biostudies/studies/E-GEOD-7947We generated RNA's and measured expression of 40000 genes using spotted cDNA microarrays from the fifty nine publicly available cell lines of the NCI Developmental Therapeutics Program's NCI60 studies and an additional set of seven cell lines for which GI50 compound sensitivity data were available. All cell lines were grown to 80% confluence in RPMI 1640 supplemented with phenol red, glutamine (2 mM) and 5% fetal calf serum. This expression data, in conjunction with the compound sensitivity data sets available from the DTP, were used to empirically determine whether gene-compound correlates of a sufficiently high correlation coefficient would have a suitable low false discovery rate to support the use of a correlative approach and these datasets for early discovery approaches for new targeted therapies. A cell type comparison design experiment design type compares cells of different type for example different cell lines. Sixty six cell lines were analysed, with no repetitions. Twelve cell lines were pooled for a common reference.biostudies-arrayexpressLabeling - For each comparative array hybridization, labeled cDNA was synthesized by reverse transcription from test cell mRNA in the presence of Cy5-dUTP, and from the reference mRNA with Cy3-dUTP, using the Superscript II reverse-transcription kit (Gibco-BRL).Growth Protocol - Cell lines were grown from the publicly available NCI DTP frozen stocks in RPMI-1640 supplemented with phenol red, glutamine (2 mM) and 5% fetal calf serum. To minimize the contribution of variations in culture conditions or cell density to differential gene expression, we grew each cell line to 80% confluence and isolated mRNA 24 hours after transfer to fresh medium.Hybridization - 65Â°C overnight in a water bath.Nucleic Acid Extraction - To minimize the contribution of variations in culture conditions or cell density to differential gene expression, we grew each cell line to 80% confluence and isolated mRNA 24 hours after transfer to fresh medium. Cells were lysed in buffer containing Protein/Rnase Degrader (Invitrogen) and messenger RNA was purified with the FastTrack 2.0 purification kit (Invitrogen).MIAME ScoreOrganizationAssays and DataProcessed DataMAGE-TAB FilesArray DesignsAssay Data Transformation - ID_REF = ID_REF<br>CH1I_MEAN = Mean feature pixel intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>CH2I_MEAN = Mean feature pixel intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale<br>CH1B_MEDIAN = The median feature background intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale; Channel: Cy3 Channel; Background<br>CH2B_MEDIAN = The median feature background intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale; Channel: Cy5 channel; Background<br>CH1D_MEAN = The mean feature pixel intensity at wavelength 532 nm with the median background subtracted.; Type: integer; Scale: linear_scale; Channel: Cy3 Channel<br>CH2D_MEAN = .The mean feature pixel intensity at wavelength 635 nm with the median background subtracted.; Type: integer; Scale: linear_scale; Channel: Cy5 channel<br>CH1I_MEDIAN = Median feature pixel intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>CH2I_MEDIAN = Median feature pixel intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale<br>CH1B_MEAN = The mean feature background intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale; Background<br>CH2B_MEAN = The mean feature background intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale; Background<br>CH1D_MEDIAN = The median feature pixel intensity at wavelength 532 nm with the median background subtracted.; Type: integer; Scale: linear_scale<br>CH2D_MEDIAN = The median feature pixel intensity at wavelength 635 nm with the median background subtracted.; Type: integer; Scale: linear_scale<br>CH1_PER_SAT = The percentage of feature pixels at wavelength 532 nm that are s:aturated.; Type: integer; Scale: linear_scale<br>CH2_PER_SAT = The percentage of feature pixels at wavelength 635 nm that are saturated.; Type: integer; Scale: linear_scale<br>CH1I_SD = The standard deviation of the feature intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale; Channel: Cy3 Channel<br>CH2I_SD = The standard deviation of the feature pixel intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale; Channel: Cy5 channel<br>CH1B_SD = The standard deviation of the feature background intensity at wavelength 532 nm.; Type: float; Scale: linear_scale; Channel: Cy3 Channel; Background<br>CH2B_SD = The standard deviation of the feature background intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale; Channel: Cy5 channel; Background<br>PERGTBCH1I_1SD = The percentage of feature pixels with intensities more than one standard deviation above the background pixel intensity, at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>PERGTBCH2I_1SD = The percentage of feature pixels with intensities more than one standard deviation above the background pixel intensity, at wavelength 635 nm.; Type: integer; Scale: linear_scale<br>PERGTBCH1I_2SD = The percentage of feature pixels with intensities more than two standard deviations above the background pixel intensity, at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>PERGTBCH2I_2SD = The percentage of feature pixels with intensities more than two standard deviations above the background pixel intensity, at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>SUM_MEAN = The sum of the arithmetic mean intensities for each wavelength, with the median background subtracted.; Type: integer; Scale: linear_scale<br>SUM_MEDIAN = The sum of the median intensities for each wavelength, with the median background subtracted.; Type: integer; Scale: linear_scale<br>RAT1_MEAN = Ratio of the arithmetic mean intensities of each spot for each wavelength, with the median background subtracted. Channel 1/Channel 2 ratio, (CH1I_MEAN - CH1B_MEDIAN)/(CH2I_MEAN - CH2B_MEDIAN) or Green/Red ratio.; Type: float; Scale: linear_scale<br>RAT2_MEAN = The ratio of the arithmetic mean intensities of each feature for each wavelength, with the median background subtracted.; Type: float; Scale: linear_scale<br>RAT2_MEDIAN = The ratio of the median intensities of each feature for each wavelength, with the median background subtracted.; Type: float; Scale: linear_scale<br>PIX_RAT2_MEAN = The geometric mean of the pixel-by-pixel ratios of pixel intensities, with the median background subtracted.; Type: float; Scale: linear_scale<br>PIX_RAT2_MEDIAN = The median of pixel-by-pixel ratios of pixel intensities, with the median background subtracted.; Type: float; Scale: linear_scale<br>RAT2_SD = The geometric standard deviation of the pixel intensity ratios.; Type: float; Scale: linear_scale<br>TOT_SPIX = The total number of feature pixels.; Type: integer; Scale: linear_scale<br>TOT_BPIX = The total number of background pixels.; Type: integer; Scale: linear_scale<br>REGR = The regression ratio of every pixel in a 2-feature-diameter circle around the center of the feature.; Type: float; Scale: linear_scale<br>CORR = The correlation between channel1 (Cy3) & Channel 2 (Cy5) pixels within the spot, and is a useful quality control parameter. Generally, high values imply better fit & good spot quality.; Type: float; Scale: linear_scale<br>DIAMETER = The diameter in um of the feature-indicator.; Type: integer; Scale: linear_scale<br>X_COORD = X-coordinate of the center of the spot-indicator associated with the spot, where (0,0) is the top left of the image.; Type: integer; Scale: linear_scale<br>Y_COORD = Y-coordinate of the center of the spot-indicator associated with the spot, where (0,0) is the top left of the image.; Type: integer; Scale: linear_scale<br>TOP = Box top: int(((centerX - radius) - Xoffset) / pixelSize).; Type: integer; Scale: linear_scale<br>BOT = Box bottom: int(((centerX + radius) - Xoffset) / pixelSize).; Type: integer; Scale: linear_scale<br>LEFT = Box left: int(((centerY - radius) - yoffset) / pixelSize).; Type: integer; Scale: linear_scale<br>RIGHT = Box right: int(((centerY + radius) - yoffset) / pixelSize); Type: integer; Scale: linear_scale<br>FLAG = The type of flag associated with a feature: -100 = user-flagged null spot; -50 = software-flagged null spot; 0 = spot valid.; Type: integer; Scale: linear_scale<br>CH2IN_MEAN = Normalized value of mean Channel 2 (usually 635 nm) intensity (CH2I_MEAN/Normalization factor).; Type: integer; Scale: linear_scale; Channel: Cy5 channel<br>CH2BN_MEDIAN = Normalized value of median Channel 2 (usually 635 nm) background (CH2B_MEDIAN/Normalization factor).; Type: integer; Scale: linear_scale; Channel: Cy5 channel; Background<br>CH2DN_MEAN = Normalized value of mean Channel 2 (usually 635 nm) intensity with normalized background subtracted (CH2IN_MEAN - CH2BN_MEDIAN).; Type: integer; Scale: linear_scale; Channel: Cy5 channel<br>RAT2N_MEAN = Type: float; Scale: linear_scale<br>CH2IN_MEDIAN = Normalized value of median Channel 2 (usually 635 nm) intensity (CH2I_MEDIAN/Normalization factor).; Type: integer; Scale: linear_scale<br>CH2DN_MEDIAN = Normalized value of median Channel 2 (usually 635 nm) intensity with normalized background subtracted (CH2IN_MEDIAN - CH2BN_MEDIAN).; Type: integer; Scale: linear_scale<br>RAT1N_MEAN = Ratio of the means of Channel 1 (usually 532 nm) intensity to normalized Channel 2 (usually 635 nm) intensity with median background subtracted (CH1D_MEAN/CH2DN_MEAN). Channel 1/Channel 2 ratio normalized or Green/Red ratio normalized.; Type: float; Scale: linear_scale<br>RAT2N_MEDIAN = Channel 2/Channel 1 ratio normalized, RAT2_MEDIAN/Normalization factor or Red/Green median ratio normalized.; Type: float; Scale: linear_scale<br>LOG_RAT2N_MEAN = Log (base 2) of the ratio of the mean of Channel 2 (usually 635 nm) to Channel 1 (usually 532 nm) [log (base 2) (RAT2N_MEAN)].; Type: float; Scale: log_base_2<br>VALUE = Log (base 2) of the ratio of the median of Channel 2 (usually 635 nm) to Channel 1 (usually 532 nm) [log (base 2) (RAT2N_MEDIAN)].; Type: float; Scale: log_base_2Feature Extraction - Array elements were manually flaged to exclude apparent problematic spots. All non-flagged array elements for which the fluorescent intensity in each channel was at least 300 units and the regression correlation between the red and green channels was greater than 0.6 were considered well measured and included in the study. Inclusion of a gene in correlative studies required that greater than 80% of measurements across all the cell lines in the study were present.Assay Data Transformation - ID_REF = ID_REF<br>CH1I_MEAN = Mean feature pixel intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>CH2I_MEAN = Mean feature pixel intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale<br>CH1B_MEDIAN = The median feature background intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale; Channel: Cy3 Channel; Background<br>CH2B_MEDIAN = The median feature background intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale; Channel: Cy5 channel; Background<br>CH1D_MEAN = The mean feature pixel intensity at wavelength 532 nm with the median background subtracted.; Type: integer; Scale: linear_scale; Channel: Cy3 Channel<br>CH2D_MEAN = .The mean feature pixel intensity at wavelength 635 nm with the median background subtracted.; Type: integer; Scale: linear_scale; Channel: Cy5 channel<br>CH1I_MEDIAN = Median feature pixel intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>CH2I_MEDIAN = Median feature pixel intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale<br>CH1B_MEAN = The mean feature background intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale; Background<br>CH2B_MEAN = The mean feature background intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale; Background<br>CH1D_MEDIAN = The median feature pixel intensity at wavelength 532 nm with the median background subtracted.; Type: integer; Scale: linear_scale<br>CH2D_MEDIAN = The median feature pixel intensity at wavelength 635 nm with the median background subtracted.; Type: integer; Scale: linear_scale<br>CH1_PER_SAT = The percentage of feature pixels at wavelength 532 nm that are saturated.; Type: integer; Scale: linear_scale<br>CH2_PER_SAT = The percentage of feature pixels at wavelength 635 nm that are saturated.; Type: integer; Scale: linear_scale<br>CH1I_SD = The standard deviation of the feature intensity at wavelength 532 nm.; Type: integer; Scale: linear_scale; Channel: Cy3 Channel<br>CH2I_SD = The standard deviation of the feature pixel intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale; Channel: Cy5 channel<br>CH1B_SD = The standard deviation of the feature background intensity at wavelength 532 nm.; Type: float; Scale: linear_scale; Channel: Cy3 Channel; Background<br>CH2B_SD = The standard deviation of the feature background intensity at wavelength 635 nm.; Type: integer; Scale: linear_scale; Channel: Cy5 channel; Background<br>PERGTBCH1I_1SD = The percentage of feature pixels with intensities more than one standard deviation above the background pixel intensity, at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>PERGTBCH2I_1SD = The percentage of feature pixels with intensities more than one standard deviation above the background pixel intensity, at wavelength 635 nm.; Type: integer; Scale: linear_scale<br>PERGTBCH1I_2SD = The percentage of feature pixels with intensities more than two standard deviations above the background pixel intensity, at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>PERGTBCH2I_2SD = The percentage of feature pixels with intensities more than two standard deviations above the background pixel intensity, at wavelength 532 nm.; Type: integer; Scale: linear_scale<br>SUM_MEAN = The sum of the arithmetic mean intensities for each wavelength, with the median background subtracted.; Type: integer; Scale: linear_scale<br>SUM_MEDIAN = The sum of the median intensities for each wavelength, with the median background subtracted.; Type: integer; Scale: linear_scale<br>RAT1_MEAN = Ratio of the arithmetic mean intensities of each spot for each wavelength, with the median background subtracted. Channel 1/Channel 2 ratio, (CH1I_MEAN - CH1B_MEDIAN)/(CH2I_MEAN - CH2B_MEDIAN) or Green/Red ratio.; Type: float; Scale: linear_scale<br>RAT2_MEAN = The ratio of the arithmetic mean intensities of each feature for each wavelength, with the median background subtracted.; Type: float; Scale: linear_scale<br>RAT2_MEDIAN = The ratio of the median intensities of each feature for each wavelength, with the median background subtracted.; Type: float; Scale: linear_scale<br>PIX_RAT2_MEAN = The geometric mean of the pixel-by-pixel ratios of pixel intensities, with the median background subtracted.; Type: float; Scale: linear_scale<br>PIX_RAT2_MEDIAN = The median of pixel-by-pixel ratios of pixel intensities, with the median background subtracted.; Type: float; Scale: linear_scale<br>RAT2_SD = The geometric standard deviation of the pixel intensity ratios.; Type: float; Scale: linear_scale<br>TOT_SPIX = The total number of feature pixels.; Type: integer; Scale: linear_scale<br>TOT_BPIX = The total number of background pixels.; Type: integer; Scale: linear_scale<br>REGR = The regression ratio of every pixel in a 2-feature-diameter circle around the center of the feature.; Type: float; Scale: linear_scale<br>CORR = The correlation between channel1 (Cy3) & Channel 2 (Cy5) pixels within the spot, and is a useful quality control parameter. Generally, high values imply better fit & good spot quality.; Type: float; Scale: linear_scale<br>DIAMETER = The diameter in um of the feature-indicator.; Type: integer; Scale: linear_scale<br>X_COORD = X-coordinate of the center of the spot-indicator associated with the spot, where (0,0) is the top left of the image.; Type: integer; Scale: linear_scale<br>Y_COORD = Y-coordinate of the center of the spot-indicator associated with the spot, where (0,0) is the top left of the image.; Type: integer; Scale: linear_scale<br>TOP = Box top: int(((centerX - radius) - Xoffset) / pixelSize).; Type: integer; Scale: linear_scale<br>BOT = Box bottom: int(((centerX + radius) - Xoffset) / pixelSize).; Type: integer; Scale: linear_scale<br>LEFT = Box left: int(((centerY - radius) - yoffset) / pixelSize).; Type: integer; Scale: linear_scale<br>RIGHT = Box right: int(((centerY + radius) - yoffset) / pixelSize); Type: integer; Scale: linear_scale<br>FLAG = The type of flag associated with a feature: -100 = user-flagged null spot; -50 = software-flagged null spot; 0 = spot valid.; Type: integer; Scale: linear_scale<br>CH2IN_MEAN = Normalized value of mean Channel 2 (usually 635 nm) intensity (CH2I_MEAN/Normalization factor).; Type: integer; Scale: linear_scale; Channel: Cy5 channel<br>CH2BN_MEDIAN = Normalized value of median Channel 2 (usually 635 nm) background (CH2B_MEDIAN/Normalization factor).; Type: integer; Scale: linear_scale; Channel: Cy5 channel; Background<br>CH2DN_MEAN = Normalized value of mean Channel 2 (usually 635 nm) intensity with normalized background subtracted (CH2IN_MEAN - CH2BN_MEDIAN).; Type: integer; Scale: linear_scale; Channel: Cy5 channel<br>RAT2N_MEAN = Type: float; Scale: linear_scale<br>CH2IN_MEDIAN = Normalized value of median Channel 2 (usually 635 nm) intensity (CH2I_MEDIAN/Normalization factor).; Type: integer; Scale: linear_scale<br>CH2DN_MEDIAN = Normalized value of median Channel 2 (usually 635 nm) intensity with normalized background subtracted (CH2IN_MEDIAN - CH2BN_MEDIAN).; Type: integer; Scale: linear_scale<br>RAT1N_MEAN = Ratio of the means of Channel 1 (usually 532 nm) intensity to normalized Channel 2 (usually 635 nm) intensity with median background subtracted (CH1D_MEAN/CH2DN_MEAN). Channel 1/Channel 2 ratio normalized or Green/Red ratio normalized.; Type: float; Scale: linear_scale<br>RAT2N_MEDIAN = Channel 2/Channel 1 ratio normalized, RAT2_MEDIAN/Normalization factor or Red/Green median ratio normalized.; Type: float; Scale: linear_scale<br>LOG_RAT2N_MEAN = Log (base 2) of the ratio of the mean of Channel 2 (usually 635 nm) to Channel 1 (usually 532 nm) [log (base 2) (RAT2N_MEAN)].; Type: float; Scale: log_base_2<br>VALUE = Log (base 2) of the ratio of the median of Channel 2 (usually 635 nm) to Channel 1 (usually 532 nm) [log (base 2) (RAT2N_MEDIAN)].; Type: float; Scale: log_base_2Image Adquisition - All arrays were scanned using GenePix 4000A microarray scanner (Axon Instruments, Union City, CA) and initial data analysis was carried out using GenePix Pro 3.0UnknownTranscriptomicsGenomicsProteomics<h4>Background</h4>The NCI has undertaken a twenty-year project to characterize compound sensitivity patterns in a selected set of sixty tumor derived cell lines. Previous studies have explored the relationship between compound sensitivity patterns to gene expression, protein expression, and DNA copy number for these same cell lines. A strong correlation between the pattern of expression of a biomarker and sensitivity to a compound could suggest a clinically interesting biological relationship between the two.<h4>Results</h4>We isolated RNA's and measured expression of 40000 genes using cDNA microarrays from the fifty-nine publicly available cell lines. Analysis of this data set in comparison with published gene expression data sets demonstrates a high degree of reproducibility in expression level measurements even using completely independent RNA preparations and array technologies. Using the fifty-nine cell lines for discovery and an additional seven cell lines for which extensive compound sensitivity data were available as a test set, we determined that gene-compound pairs with a correlation coefficient above 0.6 had a false discovery rate of approximately 5%. Large scale features of the gene expression and chemosensitivity data, such as tissue of origin and other physiological factors, did not seem to explain the majority of correlations between gene and compound patterns.<h4>Conclusion</h4>A comparison of gene expression and compound sensitivity in panels of cell lines was demonstrated to have a relatively high validation and low false discovery rate supporting the use of this approach and datasets for identifying candidate biomarkers and targeted biologically active compounds.unknown experiment typeHomo sapiensGene expression patterns within cell lines are predictive of chemosensitivity.Brian RingDoug RossDouglas RossStella ChangRing BZ, Chang S, Ring LW, Seitz RS, Ross DT64falseGene expression patterns within cell lines are predictive of chemosensitivityWe generated RNA's and measured expression of 40000 genes using spotted cDNA microarrays from the fifty nine publicly available cell lines of the NCI Developmental Therapeutics Program's NCI60 studies and an additional set of seven cell lines for which GI50 compound sensitivity data were available. All cell lines were grown to 80% confluence in RPMI 1640 supplemented with phenol red, glutamine (2 mM) and 5% fetal calf serum. This expression data, in conjunction with the compound sensitivity data sets available from the DTP, were used to empirically determine whether gene-compound correlates of a sufficiently high correlation coefficient would have a suitable low false discovery rate to support the use of a correlative approach and these datasets for early discovery approaches for new targeted therapies. A cell type comparison design experiment design type compares cells of different type for example different cell lines. Sixty six cell lines were analysed, with no repetitions. Twelve cell lines were pooled for a common reference.E-GEOD-7947GSE79471826123718261237