RNAi high throughput screening analysis.

All Answers (11) Show full discussion

• 
  Gaurab Banerjee · Sandor Proteomics

  go for a library of siRNA of your gene pathway. For validation use qPCR based taqman probes. An array format will provide you convienience. What are the cell types you are using?

  Cancel Save

  go for a library of siRNA of your gene pathway. For validation use qPCR based taqman probes. An array format will provide you convienience.
  
  What are the cell types you are using?

  Mar 29, 2012
• 
  kamariah ibrahim · National University of Malaysia

  Well, i am more concern about the statistical method use to analyze the data rather than doing rq pcr. Any suggestion?

  Cancel Save

  Well, i am more concern about the statistical method use to analyze the data rather than doing rq pcr. Any suggestion?

  Mar 30, 2012
• 
  Anastasia Eskova · Universität Heidelberg

  If with z-scores you have good separation of your positive and negative controls - why not. I think, z-scores can be used also for small sets of genes.

  Cancel Save

  If with z-scores you have good separation of your positive and negative controls - why not. I think, z-scores can be used also for small sets of genes.

  Mar 30, 2012
• 
  Pengcheng Xun · Indiana University Bloomington

  You may need to consider multiple testing issue, such as controlling FDR.

  Cancel Save

  You may need to consider multiple testing issue, such as controlling FDR.

  Mar 30, 2012
• 
  Daniel Gilbert · Friedrich-Alexander Universität Erlangen-Nürnberg

  Falschlehner C, Steinbrink S, Erdmann G, Boutros M. High-throughput RNAi screening to dissect cellular pathways: a how-to guide. Biotechnol J. 2010 Apr;5(4):368-76. PubMed PMID: 20349460.

  Cancel Save

  Falschlehner C, Steinbrink S, Erdmann G, Boutros M. High-throughput RNAi screening to dissect cellular pathways: a how-to guide. Biotechnol J. 2010 Apr;5(4):368-76. PubMed PMID: 20349460.

  Mar 30, 2012
• 
  Eugen Buehler · National Institutes of Health

  It would help if you provided a few more details. How many different RNAi reagents do you have per gene? Do you have experimental replicates?

  Cancel Save

  It would help if you provided a few more details. How many different RNAi reagents do you have per gene? Do you have experimental replicates?

  Mar 30, 2012
• 
  Namjin Chung · Bristol-Myers Squibb

  Concur with Eugen Buehler. 500 genes can be represented by 500 sample RNAi reagents, or 2000, or 5000. Regardless, try use the z-score based on median and MAD (median absolute deviation). This is often called "robust" z-score and a lot more resistant to any rogue outliers (such as pipetting errors).

  Cancel Save

  Concur with Eugen Buehler. 500 genes can be represented by 500 sample RNAi reagents, or 2000, or 5000. Regardless, try use the z-score based on median and MAD (median absolute deviation). This is often called "robust" z-score and a lot more resistant to any rogue outliers (such as pipetting errors).

  Apr 26, 2012
• 
  kamariah ibrahim · National University of Malaysia

  Dear guys, thanks for the advices. Currently I will stick with SSMD (Standardized Standard Mean Deviation) which I found more robust and result is consistent with my hypothesis that PLK1 ( positive ctrl which causing major death to my cell lines) should as well fall as well into "hits" target. By the way, I am using SmartPool sirna from dharmacon and I have triplicates for my experiment.

  Cancel Save

  Dear guys, thanks for the advices. Currently I will stick with SSMD (Standardized Standard Mean Deviation) which I found more robust and result is consistent with my hypothesis that PLK1 ( positive ctrl which causing major death to my cell lines) should as well fall as well into "hits" target. By the way, I am using SmartPool sirna from dharmacon and I have triplicates for my experiment.

  Apr 27, 2012
• 
  Eugen Buehler · National Institutes of Health

  I would not recommend SSMD for dealing with replicates. It assumes that each reagent has a unique experimental variance, and as a result tends to rank highly reagents (in this case pools) that by chance had replicates close to each other but whose median effect is significantly below that of other reagents . Simply taking the median of your replicates would be a better idea, in my opinion. What is your plan for confirmation of selected hits from your small screen? I would recommend that you use multiple (three or four) siRNAs from another vendor to confirm (do not simply use the single siRNAs from the SmartPool, as you have already tested them in the pool and they will not represent independent evidence). Also, you should plan on a relatively low confirmation rate. The performance of a single reagent is not strong evidence that the targeted gene is effecting the assay. Most of your hits (~80%) will turn out to be off-target effects.

  Cancel Save

  I would not recommend SSMD for dealing with replicates. It assumes that each reagent has a unique experimental variance, and as a result tends to rank highly reagents (in this case pools) that by chance had replicates close to each other but whose median effect is significantly below that of other reagents . Simply taking the median of your replicates would be a better idea, in my opinion. What is your plan for confirmation of selected hits from your small screen? I would recommend that you use multiple (three or four) siRNAs from another vendor to confirm (do not simply use the single siRNAs from the SmartPool, as you have already tested them in the pool and they will not represent independent evidence). Also, you should plan on a relatively low confirmation rate. The performance of a single reagent is not strong evidence that the targeted gene is effecting the assay. Most of your hits (~80%) will turn out to be off-target effects.

  Apr 27, 2012
• 
  Namjin Chung · Bristol-Myers Squibb

  So you have 150 data points (say, averages of triplicates for 150 smartpool siRNAs). If these 150 genes were not randomly chosen, I'd say any hit selection metrics used for HTS (such as z-score, SSMD, etc) are less useful. You could simply calculate averages of triplicate wells and then run unpaired t-test for each sample against your negative control (=non-targeting). Many folks use 2x changes and p <0.05 or 0.01. Even without any of these, going over 150 rows excel table won't be too bad:-)

  Cancel Save

  So you have 150 data points (say, averages of triplicates for 150 smartpool siRNAs). If these 150 genes were not randomly chosen, I'd say any hit selection metrics used for HTS (such as z-score, SSMD, etc) are less useful. You could simply calculate averages of triplicate wells and then run unpaired t-test for each sample against your negative control (=non-targeting). Many folks use 2x changes and p <0.05 or 0.01. Even without any of these, going over 150 rows excel table won't be too bad:-)

  Apr 27, 2012
• 
  kamariah ibrahim · National University of Malaysia

  i have tried using +/- kMAD analysis. how do u determine the cut off point to get hit genes? 1SD? 2SD?

  Cancel Save

  i have tried using +/- kMAD analysis. how do u determine the cut off point to get hit genes? 1SD? 2SD?

  May 18, 2012