Examination of Global 3' Bias on Microarrays
| Users of Affymetrix GeneChip® probe arrays often monitor 3' to 5' (3'/5') ratios of housekeeping genes to control for sample quality and amplification reaction efficiencies (see sidebar, "What is the 3' to 5' Ratio?"). Similar quality checks are performed during the manufacturing process of Applied Biosystems Ambion® MessageAmp™ products. The newest addition to the product line, the MessageAmp Premier Kit, is no exception. In addition to a number of other specific metrics, the quality control procedure requires that aRNA prepared using the kit meets rigorous 3'/5' signal ratios when hybridized to the GeneChip human U133A2.0 probe array, before it can be sold as an Applied Biosystems product. These array-based functional tests for quality control are among the most stringent in the industry. |
Unexpected 3'/5' Signal Ratios From the ß-actin Probe Set
During the development of the MessageAmp Premier Kit, we observed that, with decreasing RNA input, the 3'/5' signal ratios for the ß-actin probe set on Affymetrix microarrays increases. The results in Figure 1 show the increased 3'/5' ratios seen when comparing microarray data from samples prepared using 50 ng and 500 ng RNA input with the MessageAmp Premier Kit. The Affymetrix GeneChip One-Cycle Kit with 1000 ng input RNA was used as a control. Both Universal Human Reference RNA (UHRR) and Human Brain Reference RNA (HBRR) showed a similar pattern of increased 3'/5' ratios for ß-actin with lower RNA input. Conversely, 3'/5' ratios seen with the GAPDH probe set did not show an increase with the lower RNA input amount. Also notice that GAPDH 3'/5' ratios at 50 ng input RNA is most similar to the control data from aRNA prepared using the Affymetrix kit.
Figure 1. Low RNA Input May Result in Higher 3'/5' Ratios for b-actin. Human Brain Reference RNA (HBRR, Applied Biosystems) and Universal Human Reference RNA (UHRR, Stratagene) were used in RNA amplification and labeling reactions with either the Ambion® MessageAmp™ Premier Kit (50 or 500 ng RNA input amount) or the Affymetrix GeneChip® One-Cycle Kit (1000 ng RNA input amount) according to respective protocols. Array results were read and ß-actin 3'/5' ratios were calculated using GCOS software. Note that more RNA is required for the Affymetrix kit protocol than for the MessageAmp Premier Kit. JMP software was used to build plots.
Figure 1. Low RNA Input May Result in Higher 3'/5' Ratios for b-actin. Human Brain Reference RNA (HBRR, Applied Biosystems) and Universal Human Reference RNA (UHRR, Stratagene) were used in RNA amplification and labeling reactions with either the Ambion® MessageAmp™ Premier Kit (50 or 500 ng RNA input amount) or the Affymetrix GeneChip® One-Cycle Kit (1000 ng RNA input amount) according to respective protocols. Array results were read and ß-actin 3'/5' ratios were calculated using GCOS software. Note that more RNA is required for the Affymetrix kit protocol than for the MessageAmp Premier Kit. JMP software was used to build plots.
Global Assessment of 3' Bias Confirms that GAPDH is a Better Indicator than ß-actin
Figure 2 offers a different view of the data that is based on a more global assessment of 3' bias. Signal from 3'-most and 5'-most probes in the probe sets with the most complete transcript coverage, 11 probes, were used to calculate 3'/5' ratios. The distribution of these ratios was then plotted as box plots. These results indicate that samples prepared using the MessageAmp Premier Kit with 50 ng of input RNA is most similar to that prepared using the Affymetrix kit and 1000 ng of input RNA. This result is more similar to that seen with GAPDH and not ß-actin, indicating that GAPDH may be a better indicator of global 3' bias than ß-actin.
Figure 2. 3'/5' Signal Ratios for All Probe Sets. Partek GS software was used to extract probe level signals for all probe sets with n = 11 probes (RMA Background subtraction was applied without quantile normalization). 3'/5' ratios were calculated by simply dividing the 3'-most probe by the 5'-most probe for each probe set. The ratios for all probe sets were then represented as standard box plots.
Figure 2. 3'/5' Signal Ratios for All Probe Sets. Partek GS software was used to extract probe level signals for all probe sets with n = 11 probes (RMA Background subtraction was applied without quantile normalization). 3'/5' ratios were calculated by simply dividing the 3'-most probe by the 5'-most probe for each probe set. The ratios for all probe sets were then represented as standard box plots.
Conclusion
These results suggest that ß-actin may not always be an accurate indicator of global 3'/5' ratios, particularly at low RNA input amounts. Using 3'/5' ratios from the GAPDH probe set or other probe sets may provide a better indication of sample quality and reaction effectiveness than ß-actin 3'/5' ratios in microarray experiments.
Scientific Contributors:
Kelli S. Bramlett, Penn Whitley, Charmaine San Jose, Natalie Sununpong Hernandez, Bob Setterquist • Applied Biosystems Inc., Austin, TX
Scientific Contributors:
Kelli S. Bramlett, Penn Whitley, Charmaine San Jose, Natalie Sununpong Hernandez, Bob Setterquist • Applied Biosystems Inc., Austin, TX
