Limits...
Advances in genome-wide RNAi cellular screens: a case study using the Drosophila JAK/STAT pathway.

Fisher KH, Wright VM, Taylor A, Zeidler MP, Brown S - BMC Genomics (2012)

Bottom Line: Following reanalysis of the original screen data, comparisons of the two gene lists allows us to make estimates of false discovery rates in the SRSF data and to conduct an assessment of off-target effects (OTEs) associated with both libraries.We discuss the differences and similarities between the resulting data sets and examine the relative improvements in gene discovery protocols.Our work represents one of the first direct comparisons between first- and second-generation libraries and shows that modern library designs together with methodological advances have had a significant influence on genome-scale RNAi screens.

View Article: PubMed Central - HTML - PubMed

Affiliation: The MRC Centre for Developmental and Biomedical Genetics and The Department of Biomedical Science, University of Sheffield, Firth Court,Western Bank, Sheffield S10 2TN, UK.

ABSTRACT

Background: Genome-scale RNA-interference (RNAi) screens are becoming ever more common gene discovery tools. However, whilst every screen identifies interacting genes, less attention has been given to how factors such as library design and post-screening bioinformatics may be effecting the data generated.

Results: Here we present a new genome-wide RNAi screen of the Drosophila JAK/STAT signalling pathway undertaken in the Sheffield RNAi Screening Facility (SRSF). This screen was carried out using a second-generation, computationally optimised dsRNA library and analysed using current methods and bioinformatic tools. To examine advances in RNAi screening technology, we compare this screen to a biologically very similar screen undertaken in 2005 with a first-generation library. Both screens used the same cell line, reporters and experimental design, with the SRSF screen identifying 42 putative regulators of JAK/STAT signalling, 22 of which verified in a secondary screen and 16 verified with an independent probe design. Following reanalysis of the original screen data, comparisons of the two gene lists allows us to make estimates of false discovery rates in the SRSF data and to conduct an assessment of off-target effects (OTEs) associated with both libraries. We discuss the differences and similarities between the resulting data sets and examine the relative improvements in gene discovery protocols.

Conclusions: Our work represents one of the first direct comparisons between first- and second-generation libraries and shows that modern library designs together with methodological advances have had a significant influence on genome-scale RNAi screens.

Show MeSH

Related in: MedlinePlus

Visualisation of whole genome data is required for error identification. (A-B) Box and whisker plots show the averages and variance both of the whole data set (All Samples) as well as the controls used in the HFA screen and SRSF screens. GFP RNAi resulted in only a weak decrease in signalling, due to the high levels of Upd-GFP produced in the transfection. Failure of notches to overlap suggests that the two medians are significantly different. Y-axis shows Z-scores. (C-F) Heat maps representing Z-scores of FL/RL normalised values for the HFA screen (C and E) and the SRSF screen (D and F), in unfiltered format (C and D) and after filtering of controls and errors (E and F). As shown in the key, blues represent a decrease in pathway activity while reds indicate an increase.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3526451&req=5

Figure 2: Visualisation of whole genome data is required for error identification. (A-B) Box and whisker plots show the averages and variance both of the whole data set (All Samples) as well as the controls used in the HFA screen and SRSF screens. GFP RNAi resulted in only a weak decrease in signalling, due to the high levels of Upd-GFP produced in the transfection. Failure of notches to overlap suggests that the two medians are significantly different. Y-axis shows Z-scores. (C-F) Heat maps representing Z-scores of FL/RL normalised values for the HFA screen (C and E) and the SRSF screen (D and F), in unfiltered format (C and D) and after filtering of controls and errors (E and F). As shown in the key, blues represent a decrease in pathway activity while reds indicate an increase.

Mentions: In the first instance, controls included on each screening plate were assessed to determine the biological effectiveness of both screens (Figure 2A and 2B). In both screens positive controls known to be required for JAK/STAT signalling are recovered. Furthermore, the additional space available for controls in the SRSFv1 library also allows for technical controls (RLuc, which targets Renilla luciferase mRNA and thus skews the FL/RL ratio, and GFP, which targets the Upd-GFP mRNA) and the negative control (C.elegans dsRNA ZK686.3) to be added (Figure 2B). In addition, the effect of inter-plate and inter-replicate effects were also assessed using box and whisker graphs plotted for each plate (Additional file 2). When compared in this manner, the SRSFv1 data appears to be highly consistent - with the exception of plates 38 and 53 in replicate 3 (Additional file 2E, asterisks). By contrast, the HFA data shows considerably more variation between plates, a characteristic that often spans both replicates (Additional file 2A and 2B, asterisks).


Advances in genome-wide RNAi cellular screens: a case study using the Drosophila JAK/STAT pathway.

Fisher KH, Wright VM, Taylor A, Zeidler MP, Brown S - BMC Genomics (2012)

Visualisation of whole genome data is required for error identification. (A-B) Box and whisker plots show the averages and variance both of the whole data set (All Samples) as well as the controls used in the HFA screen and SRSF screens. GFP RNAi resulted in only a weak decrease in signalling, due to the high levels of Upd-GFP produced in the transfection. Failure of notches to overlap suggests that the two medians are significantly different. Y-axis shows Z-scores. (C-F) Heat maps representing Z-scores of FL/RL normalised values for the HFA screen (C and E) and the SRSF screen (D and F), in unfiltered format (C and D) and after filtering of controls and errors (E and F). As shown in the key, blues represent a decrease in pathway activity while reds indicate an increase.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3526451&req=5

Figure 2: Visualisation of whole genome data is required for error identification. (A-B) Box and whisker plots show the averages and variance both of the whole data set (All Samples) as well as the controls used in the HFA screen and SRSF screens. GFP RNAi resulted in only a weak decrease in signalling, due to the high levels of Upd-GFP produced in the transfection. Failure of notches to overlap suggests that the two medians are significantly different. Y-axis shows Z-scores. (C-F) Heat maps representing Z-scores of FL/RL normalised values for the HFA screen (C and E) and the SRSF screen (D and F), in unfiltered format (C and D) and after filtering of controls and errors (E and F). As shown in the key, blues represent a decrease in pathway activity while reds indicate an increase.
Mentions: In the first instance, controls included on each screening plate were assessed to determine the biological effectiveness of both screens (Figure 2A and 2B). In both screens positive controls known to be required for JAK/STAT signalling are recovered. Furthermore, the additional space available for controls in the SRSFv1 library also allows for technical controls (RLuc, which targets Renilla luciferase mRNA and thus skews the FL/RL ratio, and GFP, which targets the Upd-GFP mRNA) and the negative control (C.elegans dsRNA ZK686.3) to be added (Figure 2B). In addition, the effect of inter-plate and inter-replicate effects were also assessed using box and whisker graphs plotted for each plate (Additional file 2). When compared in this manner, the SRSFv1 data appears to be highly consistent - with the exception of plates 38 and 53 in replicate 3 (Additional file 2E, asterisks). By contrast, the HFA data shows considerably more variation between plates, a characteristic that often spans both replicates (Additional file 2A and 2B, asterisks).

Bottom Line: Following reanalysis of the original screen data, comparisons of the two gene lists allows us to make estimates of false discovery rates in the SRSF data and to conduct an assessment of off-target effects (OTEs) associated with both libraries.We discuss the differences and similarities between the resulting data sets and examine the relative improvements in gene discovery protocols.Our work represents one of the first direct comparisons between first- and second-generation libraries and shows that modern library designs together with methodological advances have had a significant influence on genome-scale RNAi screens.

View Article: PubMed Central - HTML - PubMed

Affiliation: The MRC Centre for Developmental and Biomedical Genetics and The Department of Biomedical Science, University of Sheffield, Firth Court,Western Bank, Sheffield S10 2TN, UK.

ABSTRACT

Background: Genome-scale RNA-interference (RNAi) screens are becoming ever more common gene discovery tools. However, whilst every screen identifies interacting genes, less attention has been given to how factors such as library design and post-screening bioinformatics may be effecting the data generated.

Results: Here we present a new genome-wide RNAi screen of the Drosophila JAK/STAT signalling pathway undertaken in the Sheffield RNAi Screening Facility (SRSF). This screen was carried out using a second-generation, computationally optimised dsRNA library and analysed using current methods and bioinformatic tools. To examine advances in RNAi screening technology, we compare this screen to a biologically very similar screen undertaken in 2005 with a first-generation library. Both screens used the same cell line, reporters and experimental design, with the SRSF screen identifying 42 putative regulators of JAK/STAT signalling, 22 of which verified in a secondary screen and 16 verified with an independent probe design. Following reanalysis of the original screen data, comparisons of the two gene lists allows us to make estimates of false discovery rates in the SRSF data and to conduct an assessment of off-target effects (OTEs) associated with both libraries. We discuss the differences and similarities between the resulting data sets and examine the relative improvements in gene discovery protocols.

Conclusions: Our work represents one of the first direct comparisons between first- and second-generation libraries and shows that modern library designs together with methodological advances have had a significant influence on genome-scale RNAi screens.

Show MeSH
Related in: MedlinePlus