NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Sittampalam GS, Coussens NP, Brimacombe K, et al., editors. Assay Guidance Manual [Internet]. Bethesda (MD): Eli Lilly & Company and the National Center for Advancing Translational Sciences; 2004-.

Cover of Assay Guidance Manual

Assay Guidance Manual [Internet].

Show details

Cell-Based RNAi Assay Development for HTS

, , , , , and .

Author Information

Published ; Last Update: May 1, 2013.


Gene silencing through RNA interference (RNAi) has become a powerful tool for understanding gene function. RNAi screens are primarily conducted using synthetic small interfering RNA (siRNA) or plasmid-encoded short hairpin RNA (shRNA). In this chapter, some considerations for design, optimization, validation, analysis and hit selection criteria in RNAi screens are discussed. A special emphasis is placed on pitfalls associated with off-target effects, which represent a primary limitation to the successful application of this technology.


RNA interference (RNAi) is a gene silencing mechanism initiated by short double-stranded RNA (dsRNA) of ~21nt in length (for a recent review see 1). Two major classes of dsRNAs harness this pathway for post-transcriptional gene regulation, including siRNAs and microRNAs (miRNAs). siRNAs direct the cleavage of mRNA transcripts that contain full sequence complementarity. Cleavage is mediated by a single strand of the siRNA duplex termed the guide strand, after loading into the RNA-induced silencing complex (RISC). Notably, the documented occurrence of naturally occurring cleavage complexes is not common in mammalian cells. Rather, it is miRNAs that use the innate RNAi machinery. miRNAs interact with transcripts possessing partial complementarity, primarily within target 3′ untranslated regions (UTRs), resulting in transcript degradation and/or translation inhibition (Figure 1).

Figure 1. . Simplified schematic of RNAi in mammalian cells.

Figure 1.

Simplified schematic of RNAi in mammalian cells. RNAi in mammalian cells is primarily mediated by endogenous miRNAs. miRNAs are expressed as primary hairpin-containing transcripts that are processed in the nucleus and cytoplasm to yield mature miRNA duplexes (more...)

Experimentally, the ability to harness the RNAi pathway through the use of siRNAs/shRNAs (Figure 2) has paved the way for genome-wide high throughput screens. Many large-scale RNAi screens have been reported. Common variations include drug modifier screens, which combine the use of RNAi and a drug to identify genes that affect drug response, viability screens to look for vulnerabilities within specific cellular backgrounds, pathway reporter assays, pathogen-host screens to look for genes that affect pathogen spread and host response, and image-based phenotypic screens to report on genes associated with a wide variety of processes, including protein localization and disease-specific phenotypes. Many reviews cover the RNAi biology, experimental parameters and considerations for performing screens (for example, 2-11). See table 1 for a brief summary of the differences between siRNA and shRNA reagents.

Figure 2. . siRNAs and shRNAs harness the RNAi pathway for loss-of-function studies.

Figure 2.

siRNAs and shRNAs harness the RNAi pathway for loss-of-function studies. shRNAs are encoded by plasmids and processed much like miRNAs to yield mature duplexes of ~22 nt in length. Alternatively, synthetic siRNAs can be introduced directly into the cell (more...)

Table 1:

Table 1:

Comparison of siRNA and shRNA-lentivirus

In the following sections, we have compiled our experience around RNAi-based LOF screens in mammalian cells to offer a few guidelines on best practices. As with any technology, this chapter will benefit from growing expertise and improvements in technology and methods.

Off-Target Effects

The ability of siRNAs/shRNAs to knockdown intended targets while minimizing or controlling for off-target effects (OTEs) is critical for the meaningful interpretation of RNAi screens. Off-target effects arise from mechanisms that can be either independent or dependent upon the siRNA/shRNA sequence. Sequence-independent effects can relate to experimental conditions (e.g., transfection reagents), inhibition of endogenous miRNA activity, or stimulation of pathways associated with the immune response. Sequence-dependent effects primarily concern the unintentional silencing of targets sharing partial complementarity with RNAi effector molecules through miRNA-like interactions. There are a number of approaches toward controlling and accounting for both types of off-target effects (discussed below).

Sequence-Dependent Off-Target Effects: Interactions between siRNAs/shRNAs and non-targeted mRNAs

Although RNAi reagents can cause sequence-independent effects, the primary source of trouble for RNAi screeners are sequence-dependent off-target effects. Off-target effects originate from partial complementarity between RNAi effectors and off-target transcripts, in much the same way as those exhibited by endogenous miRNAs. In fact, like miRNA targets, off-targeted transcripts are enriched in those containing perfect pairing between their 3′ UTRs and hexamer (nts 2–7) and heptamer (nts 2–8) sequences within 5′ ends of RNAi effectors (12,13). These stretches of sequence are known as “seed sequences”. Some studies have found these effects to be non-titratable, with dose responses mirroring that of on-target transcripts (14). Others have found these effects to be concentration-dependent, whereby the use of low siRNA concentrations can significantly mitigate off-target interactions (15). Sequence-dependent off-target effects can have profound consequences. For example, Lin and colleagues determined that the top three “hits” from a siRNA-based screen for targets affecting the hypoxia-related HIF-1 pathway resulted from off-target effects (16). For two of these three “hits,” activity could be traced to interactions within the 3′ UTR of HIF-1A itself. Additionally, Schultz and coworkers found that all active siRNAs in a TGF-β assay reduced TGFBR1 and TGFBR2 (17).

How bad is the problem with seed-driven OTEs? It has been estimated that in a genome-wide screen using 4 siRNAs per gene, and an estimated 20 true positives in the assay, that 3,362 off-target genes would score with 1 active siRNA, 259 would score with 2 of 4 active siRNAs, and 9 would score with 3 of 4 (18). This is a sobering estimate given that the vast majority of published RNAi screen are conducted with only a single reagent per gene (pool of 4 siRNAs), and the typical bar for follow-up validation is to require that 2 members of the pool, or sometimes even 1, exhibit activity. Those approaches seem insufficient, and lead to published hit lists that are loaded with false positives, especially in cases where a simple laundry list of actives from the primary screen are presented. An additional illustration of this problem can be found in recently published host-virus screens. For example, a meta-analysis of 3 genome-wide siRNA screens conducted in human cells to look for host genes associated with HIV revealed strikingly little overlap (19). Notably, only 3 genes were called in all 3 screens, and the pair-wise comparison of any two screens revealed only 3%-6% overlap. However, pathway analysis revealed greater similarities between the screens, and certainly false negatives (e.g, arising from reagent deficiencies) and differences in experimental set up (e.g., cell lines and assay endpoints) are significant contributors to the lack of agreement. However, even when comparing different siRNA libraries under the same exact experimental conditions there is virtually no correlation (Figure 3). In fact, the correlation between siRNAs having the same seed is much greater than siRNAs designed to target the same gene (20, Figure 4), emphasizing the prevalence and impact of seed-driven OTEs in RNAi screen data.

Figure 3. . Comparing different siRNA reagents under the same exact experimental conditions.

Figure 3.

Comparing different siRNA reagents under the same exact experimental conditions. Two different screens show very little correlation between different siRNAs desigend to target the same gene.

Figure 4. . The correlation between siRNAs having the same seed is much greater than siRNAs designed to target the same gene (20).

Figure 4.

The correlation between siRNAs having the same seed is much greater than siRNAs designed to target the same gene (20). This is clear evidence that seed-dependent OTEs are the primary reason for a lack of agreement between siRNAs designed to target the (more...)

There are a number of ways to help minimize the impact of sequence-dependent off-target effects. For starters, an attempt should be made to use siRNAs at relatively low concentrations. Early studies used siRNAs at ≥ 100nM, but it should be possible to routinely use them at 10 nM – 50 nM without loss of on-target potency. Other ways to reduce OTEs, relate to siRNA design features and chemical modifications. For example, siRNAs are now commercially available with chemical modifications to the passenger strand, which eliminate their loading into RISC, and the subsequent off-target effects that may result. Redundancy (the use of multiple reagents per gene) is another way to minimize the impact of off-target effects by requiring multiple active reagents per gene for that gene to be considered a candidate active. There are also informatic approaches to identify and even interpret off-target effects within RNAi screens. These will be discussed in more detail below. Despite all of these considerations, the occurrence of sequence-dependent off-target effects is unavoidable.

Loss-of-Function Screens Using siRNA

siRNA Reagents

The following choices of reagents need to be made prior to running any screens.

  • Scale: Focused libraries (pathway collection, gene family, disease-specific library, etc.), druggable genome, or genome-wide. A variety of vendors offer these reagents (e.g., Qiagen, Dharmacon, Ambion, Sigma).
  • Format: Some vendors provide pools of siRNAs against a given gene in an effort to guarantee knockdown. Others provide libraries in a single siRNA per well format. Recently, a variety of chemically modified siRNAs have become available. These modifications reduce off-target effects, especially arising from the passenger strand, and should be used.
  • In light of the issue with OTEs, the use of multiple reagents per gene in a screen will increase the chances of identifying true positives. This is illustrated in the meta-analysis of HIV screen for example, in which genes called in 2 or 3 of the screens were more enriched in relevant pathways (19). Screening multiple reagents per gene can be a more expensive option and increase the scale of the screen. However, it is common practice to screen one reagent per gene in duplicate or triplicate. Given that the majority of variance arises from false negatives and positives (see figures 3 and 4) and that the correlation between replicates in a well-optimized assay can be quite high (see the pilot screens section below), it would seem wise to invest more in redundancy than replicates, if a choice must be made (i.e., 3 different reagents per gene can be screened for the same cost as 1 reagent per gene in triplicate), provided that an assay has been demonstrated to be highly reproducible.
  • Negative and positive controls: Negative and positive controls should be embedded in every assay plate. Negative controls are available from a number of vendors and are designed to lack homology with known transcripts. Positive controls should affect the assay under investigation (e.g., block the spread of virus in a virus assay). In cases where a good positive control does not exist, siRNAs should be chosen to at least report on the quality of transfection (e.g., lethal siRNAs that target essential genes or siRNAs that target the reporter used in a given assay, like GFP or luciferase). It is also important to note that negative controls are most likely not truly negative in any given assay.

Assay Optimization

Optimization needs to be done for all screens. Table 2 lists some important parameters for consideration in RNAi optimization.

Table 2:

Table 2:

Important parameters in RNAi-assay optimization

A few essential parameters (and their purpose) are worth highlighting:

  • General guidelines for cell-based assays such as growth media, seeding density, growth rate, incubation time, etc. can be found in the Cell-Based Assay section in this manual.
  • Timing: Typical siRNA screen range from 48 h to 120 h. siRNA can reach maximal silencing of mRNA transcripts within 12h – 24h, but concomitant loss or protein will depend on protein half-life. If stimuli is to be added (e.g., drug or virus), it is typical to add it 48 h – 72 h post-transfection to ensure protein knockdown for a majority of genes prior to treatment. It is also important to remember that loss of silencing will begin to occur around ~96 h, so careful considerations must be made when designing an assay.
  • Transfection efficiency (below are some of the most important parameters for RNAi optimization, with reverse transfection being the preferred method for screening). See the “siRNA Transfection Optimization Experiments” section below.
    • Cell seeding density (e.g., for a viability-based experiment, you would not want to reach confluence prior to the assay endpoint).
    • Choice of transfection reagent and the amount
    • siRNA concentration (typically 10nM – 50nM)
  • Determination of KD efficiency along with transfection efficiency should constitute an essential part of assay development and optimization. The extent of KD can be determined by qRT-PCR quantification of target transcript level after si/shRNA treatment. Transfection efficiency can be gauged with positive controls where a phenotypic effect (such as cell killing or reporter gene knockdown) is observed. It is ideal to use a positive control that is sensitive to knockdown efficiency, meaning that the effects are not observed under suboptimal transfection conditions. The use of such controls would also allow evaluation of both transfection and KD efficiency after a large scale screening to check performance.
  • Choice of controls
    • Positive controls: cells with gene specific si/shRNAs transfected that will result in a significant change to the assay readout. For example si/shRNAs targeting UBB or PLK1 can be used as positive controls in cell proliferation or apoptosis assays. These controls can be informative in evaluating transfection efficiency and KD efficiency.
    • Negative controls: cells with non-silencing si/shRNAs (NS), also known as non-targeting control (NTC), transfected but without significant effect on the assay readout. It is also recommended to include the following as negative controls, although there may not be enough real estate in screen plates for these to be included throughout the screen:
      • Non-transfected (NT) cells: cultured cells only, without transfection/infection
      • Mock-transfected (MT) cells: cells with transfection reagent only, without si/shRNA
    • It is important to verify that transfection conditions do not significant alter the assay. For example, drug efficacy should be the same in cells transfected with negative control siRNA versus NT cells in a drug modifier screen. Similarly, cell transfected with negative control siRNA should not respond differently to virus than NT cells.
  • Assay z’-factors: After identifying optimal transfection conditions, it is important to evaluate the assay z’-factor to understand if the assay signal window and variation are at an acceptable level. Although, a factor of 0.5 is widely accepted for small molecule screens, lower assay z’-factors are generally accepted and expected for RNAi screens.

siRNA Transfection Optimization Experiments

A convenient first step in optimizing siRNA transfection conditions is to use viability assays with lethal control siRNAs. Two examples of a 96-well plate layout for transfection optimization are shown below (Figure 5). The bottom line is that a number of variables, including cell seeding density and transfection reagent identity/concentration, should be assessed.

Figure 5. . siRNA optimization plate layout examples.

Figure 5.

siRNA optimization plate layout examples. (A) has a wider range of transfection reagent concentrations to judge, given a fixed siRNA concentration (the edge wells are intentionally left blank); while in (B) the concentrations of siRNA and transfection (more...)

The optimal condition is determined mainly by:


Negative controls should closed mimic NT; while the positive controls should achieve maximal effect as determined by the optimization experiments. Knockdown efficiency using the best condition should be verified by real-time PCR with previously validated siRNAs.


Controls should not exhibit high variance, which would indicate significant variation in transfection efficiency.


Performing experiments in clear bottom plates is also an excellent way to visualize transfection efficiency, as virtually all cells in a given well should be visibly affected by transfection with lethal controls (e.g., cell rounding).


After identifying potential conditions via lethal control experiments, those conditions should still be tested with assay-specific positive controls.

Pilot Screens

Pilot screens should be performed when conducting large-scale siRNA screens. Pilot screens will help inform on assay reproducibility and data distribution. Pilot screens can be conducted with defined subsets of the genome-wide library such as the kinome. Replicate screens conducted at the same time should exhibit large correlation (r2~0.8) and even replicate screens conducted at different times (e.g., weeks or day apart) should be highly reproducible (see Figure 6). Pilot screen will also indicate how the data will be distributed in the larger screening campaign (e.g., normal, log-normal, or non-normal) and help indicate problems. For example, a very active screen, even if reproducible, may make it impossible to find anything of true significance in the assay, if one determines significance based on the screen population, and not a comparison to a single negative control (see hit selection below). Pilot screens can also indicate edge effects and other assay artifacts (Figure 7).

Figure 6. . An example of a highly reproducible assay.

Figure 6.

An example of a highly reproducible assay. Pilot kinome screens conducted on separate occasions indicate highly reproducible assay conditions - a prerequisite for conducting an RNAi screen. Using siRNA sequences in pilot screens that are also represented (more...)

Figure 7. . Pilot screens can indicate technical problems such as positional biases.

Figure 7.

Pilot screens can indicate technical problems such as positional biases. Here, the assay signal is clearly biased toward the middle of the plate. This also emphasizes the value of data visualization.

RNAi Screen Design and Quality Control

Plate Design: When designing the plates one should consider including a sufficient number of control replicates to help evaluate data quality. The number of wells for each type of control within the plate should be ≥ 4 for 96-well plates (preferably 8 wells); or ≥ 8 for 384-well plates (preferably 16 wells). See Figure 8 for a sample screen plate layout.

Figure 8. . Experimental plate layout example shown in 384-well format.

Figure 8.

Experimental plate layout example shown in 384-well format.

Note: Plate layouts and available wells may limit the incorporation of MT or NT controls. However, control plates should always be included in screen batches to show that negative control transfected cells respond similar to MT and NT.

Replicates and Redundancy: Ideally, one would incorporate numerous, different reagents per gene (redundancy) and biological replicates. Given the issues with siRNA screening, it has been suggested that 4-6 reagents per gene be screened in the primary campaign (21). Unfortunately, many off-the-shelf libraries, whether whole genome or focused, come with one reagent per gene (pool) or 3 singles per gene, requiring multiple expensive purchases to achieve that threshold. Notably, it is much easier to acquire ≥4 reagents per gene when constructing a custom library in conjunction with a vendor. Performing replicate screens can be another approach to honing in on the most reproducible actives. For example, in the meta-analysis of HIV screens by Bushman and colleagues (19), simulations using the estimated variance for one of the siRNA screens predicted that of the top 300 hits only half would be shared between two replicate screens. Increasing the number of replicates to ten would increase the overlap to 240 genes. However, the cost of running multiple reagents per gene ten times is prohibitive, requiring enough reagents (e.g., plates, tips, cell culture, transfection reagent, assay reagent, etc.) to run ~2400 384-well plates. Given cost and throughput realities, choices must be made. For a highly reproducible assay, as determined in the pilot phase of a screening campaign, it is a given that most of the variance will come from a lack of agreement between reagents designed to target the same gene (false positives and negatives) and one may wish to devote more resources to redundancy as compared to replicates. If pilot screens indicate a high variance and very little overlap between replicates, then many replicates or a complete overhaul of the assay will be required. For smaller, focused screens, it should be practical to run multiple reagents per gene in duplicate and beyond if necessary (the correlation between replicates can dictate the ideal number of replicates).

Quality Control

Uniformity: Uniformity within-plate or from plate to plate is also a key factor to check for quality control. Heat maps are recommended to visualize each screen plate as they help to identify geometric effects due to experimental errors or systematic problems. Section II of this manual has guidelines on within-plate variation evaluation. Given the steps involved and the cell-based nature of RNAi experiments, CVs (coefficient of variation) ranging from ~15% to > 30% of the sample population are common. A scatter plot of pate-wise CVs versus the plate index may reveal plate-to-plate differences. In general most plates would be expected to yield similar CVs, but undoubtedly there will be biased plates in a given library given that libraries are not always randomly distributed (e.g., a plate may be rich in ribosomal proteins). If there are outlier plates in terms of CV, it will be important to inspect the source of those differences on a case by case basis. A replicate of that plate will ultimately determine if the aberrant CV is in fact a reflection of the biology occurring on the plate, or an assay artifact. Additionally, B-score normalization may help to minimize systematic row and column variations.

Control Variation: Scatter plots of common control wells across plates also help to evaluate plate-plate variation (Figure 9).

Figure 9. . Scatter plots from two exemplar screens.

Figure 9.

Scatter plots from two exemplar screens. Plots A and B show plate to plate variability as assessed by controls (different colors for different controls). It can be seen that the signals (Y-axis) for each kind of control are similar from plate to plate (more...)

Assay Z’-factors: In HTS, a scatter plot of Z prime factors for every plate versus the plate index will reveal the plate-plate differences and may help to troubleshoot any existing problems, or flag plates for redo.

Transfection / Infection Efficiency: In many cases, the ratios between the negative control and positive control will inform on transfection/infection efficiency. For example, in a cell viability assay, the ratio of the potent positive control versus the negative controls can ideally be <5%, which can be interpreted as high (> 95%) transfection/infection efficiency. While there is no theoretically defined threshold value, often these ratios will depend on the type of assays and potency of controls. In some assays, comparing positive and negative controls will not obviously inform on transfection efficiency or homogeneity. In those cases, it may be advisable to run a control plate with siRNAs that will inform upon efficiency and heterogeneity of transfection.

Replicates: To evaluate the reproducibility between replicate plates (or within plate replicates), one can use Lin’s Concordance Correlation Coefficient (Lin’s CCC 22, 23), Bland-Altman test (24) and Pearson correlation coefficients. Please consult your statistician for the most applicable method.

siRNA Hit Selection

Normalization and Hit Selection: Data is typically normalized to controls (e.g., the median of the negative control wells on each plate) or median plate activity (although the presence of biased plates may make this less appropriate). Although normalizing to the plate median may initially seem more attractive than normalizing to a negative control (which could have assay activity due to off-target effects), this will cause problems during follow-up experiments. After selections are made for follow-up, any validation experiments will be biased as will the plate median, making it impossible to compare result between primary and secondary experiments. Therefore, we consider it preferable to control for plate to plate variation by normalizing to a negative control. Please consult your statistician for most suitable method. Hit selection in large-scale siRNA screens is usually performed by converting normalized values into z-scores, or MAD-based z-scores, and then selecting actives that cross a selected threshold (e.g., 2 z-scores or MAD scores from the screen median). To apply these types of methods it is important that the data be normally distributed. Comparison to negative control using a specified cutoff with statistical and/or biological significance (e.g. > 50% loss of cell viability) may be more appropriate for smaller screens. However, using cutoffs based on departure from negative control for larger screens may lead to too many “hits”. For example, it may be historically accepted that a 30% reduction in the spread of a particular virus is biological relevant and significant. However, a 30% reduction may only score one standard deviation from the screen median, meaning that ~15% of a normally distributed population would be considered active. Alternatively, the biologically significant reduction of 30% may represent a very significant departure from the screen median, and be a very appropriate threshold. For smaller screens with large number of replicates, statistical tests (such as two sample t-test) can be used with appropriate multiple testing correction (e.g. Tukey’s, or Dunnett) if necessary.

Hits selected from screens using reagent redundancy are typically restricted to those genes with multiple active reagents. For example, one could require that 2 or more of 4 siRNAs total cross a specified threshold. A corresponding p-value and FDR can also be associated with those criteria. Similarly, redundant siRNA analysis (RSA) considers the activity of each siRNA for a given gene in an assay and generates a corresponding rank and p-value (21). Other non-parametric methods like sum of ranks can also be employed. Consult your statistician, and be sure that hits selected for follow-up are actually grounded in the screen data.

Primary screen data can and should be filtered for off-target effects. For example, by screening a large library of non-pooled siRNAs, one can analyze the data for biased seed sequences (20, Figure 10). siRNAs containing these seeds can be demoted in hit selection. Similarly, the top active siRNAs can be filtered for those containing known miRNA seed sequences with the assumption that these seeds will be highly promiscuous in terms of off-target profiles. In addition to flagging potentially biased siRNAs, there are also tools to even interpret the underlying OTEs (25, 26).

Figure 10. . Screening a large library of non-pooled siRNAs enables determination of biased seed sequences (20).

Figure 10.

Screening a large library of non-pooled siRNAs enables determination of biased seed sequences (20). Two siRNAs targeting SCAMP5 appear to significantly down-regulate the assay response. However, siRNAs containing the same hexamer sequences exhibit a clear (more...)

Pathway analysis can also reveal enrichments in the data and prioritize hits for follow-up. A variety of commercial and open software are available. A potential caveat is that this type of analysis is biased towards well-annotated genes.

Follow-up Assays

The detailed follow up plan for hits identified in a screen would depend on the nature of the investigation and the goal(s) of the study. That said, a few general suggestions are described below.

  • Test additional siRNAs for targets of interest. These siRNAs should constitute different sequences than those in the primary screen. It has been traditionally accepted that 2 active siRNAs constitutes a validated active, but given the estimates described above, and the continued reports describing lack of correlation between published screens using similar systems, 2 active siRNAs seems to be an inadequate standard. However, any candidate can be examined in additional follow-up experiments for more rigorous validation.
  • Target gene KD can be validated. Knockdown can be measured by QPCR over 24h or 48h. Best practices for carrying out QPCR experiments can be found elsewhere in the current or future versions of the QB manual. siRNA efficacy can also be assessed by Western blotting or immunofluorescence when possible. Clearly, a siRNA that yields a phenotype, but does not yield KD is a false positive. However, demonstrating KD does not prove that the observed phenotype is due to the on-target knockdown of that gene (i.e., an siRNA that is effective against its target has no bearing on its ability to down-regulate other transcripts and cannot be interpreted as validation). Furthermore, the number of siRNAs that have no effect on their intended target is relatively small, making it unlikely that QPCR will be a cost effective method for eliminating false-positives.
  • Rescue: The current gold standard in RNAi hit validation is rescue of phenotypes by introducing siRNA-resistant cDNA. Another approach is to knockout the gene of interest by using TALEN or similar technologies. This knockout should recapitulate the siRNA-induced phenotype and can also be rescued by subsequent re-expression of the gene via cDNA. Although these experiments can be excellent validation, they can also be technically challenging and suffer from their own pitfalls.
  • Test control siRNAs that retain the seed region of the original siRNA, but not its ability to cleave the transcript of interest. Recent reports have described control siRNAs where the seed region is maintained, but bases 9-11 are altered (27). The intent is to maintain the seed-driven off-target effects of a given siRNA, while eliminating its on-target effect. This initial study showed promise in separating false from true positives.

Secondary assays: This is recommended to eliminate assay artifacts and characterize target biology in more detail. Therefore, the exact nature of the assay may differ as a function of target pathway, biological process and disease biology. Validated high-content assays maybe particularly useful in this regard. These are described elsewhere in the QB manual.

RNAi Synthetic Lethality Screens

Synthetic Lethality Screens

A variation of a LOF screen is a synthetic lethality (or, synthetic lethal) screen which combines the use of RNAi and a drug (at single concentration or multiple concentrations) to identify knockdown events that would modulate drug response such as sensitizers that enhance drug effect. This offers a powerful approach to identify genetic determinants of drug response, especially in cancer. Most of the assay optimization and follow-up assays for si/shRNA described in part B apply here. The extra optimization and differences in data analysis will be discussed below.

Assay Optimization

In synthetic lethality screens, the incubation time of the drug, its potency and stability also need to be evaluated. Drug dose and time response (DDTR) experiments can be carried out to optimize these conditions in either 96-well or 384-well plates. For instance, in a 96-well plate, 10-point drug dilutions with NT and negative controls (at fixed si/shRNA concentration) can be applied along each row excluding the two edge columns (Figure 11). Assay readouts need to be monitored over a period of time, say from Day 1 to Day 6. The result of such an experiment is mentioned below.

Figure 11. . 96-well plate layout for DDTR.

Figure 11.

96-well plate layout for DDTR. NT: non-transfected; NS: non-silencing siRNA (negative control)

Example: A DDTR experiment for one drug was done from Day 1 to Day 6 to determine the appropriate drug incubation time for subsequent siRNA synthetic lethality screens. Data from Day 3 to Day 6 in Figure 8 (data from Day 1 and Day 2 data was not informative for curve fitting). Sigmoidal dose response curves of NT and NS were obtained from the experimental plates designed as above and IC50 values were estimated for each day. From Figure 12, we can see that:

Figure 12. . DDTR example of one experiment with four replicates from Day 1 to Day 6 (Data from Day 1 and 2 was not for curve fitting and is not shown); black solid lines and solid points are for NT, red dash lines and bullet points for NS.

Figure 12.

DDTR example of one experiment with four replicates from Day 1 to Day 6 (Data from Day 1 and 2 was not for curve fitting and is not shown); black solid lines and solid points are for NT, red dash lines and bullet points for NS.

  • NS and NT produce almost exactly the same dose response curves over the various concentrations sampled
  • IC50 values of the drug (either from NT or NS curves, see vertical drop lines in Figure 12) tend to stabilize from Day 4 (for NT, Day 3: 44.59nM, Day 4: 35.78nM, Day 5: 36.54nM, Day 6: 31.77nM)
  • Signal window between the zero concentration and the highest concentration of the drug tends to stabilize from Day 5 (Z prime factor calculated using NT at concentration zero and 200 nM: Day 3-0.42; Day 4-0.89; Day 5-0.73; Day 6-0.73).

Therefore 5 days of drug treatment would be recommended.

Design of Synthetic Lethality Screens

There are two main designs for synthetic lethality screens: single and multiple concentrations of drug. The hit selection strategy will vary accordingly.

  • Single-concentration experiment - Typically drug concentrations less than the IC50 are chosen (e.g. IC10 and/or IC30). At each point including zero, we recommend at least three replicates (may reduce to duplicates in a high-throughput screen).
  • Multiple-concentration experiment - A full dose response curve of the drug is used. We recommend 7 doses with duplicates as a minimum. For larger scale screens where number of points and replicates are an issue, we would suggest increased dose points, provided they are chosen carefully to cover the full range of dose response.
    Note: Several advantages exist with a RNAi synthetic lethality screen run with multiple concentrations. Non-linear curve fitting to identify biologically more relevant hits that demonstrate a ‘shift’ in DDR is made possible. Replicates are not as major an issue and achieving exact dose effect is not a concern due to curve fitting. In our experience, it is likely to produce more robust screen actives (less false-positives) and reduce follow-up steps.

In synthetic lethality screens, other necessary considerations are:

  • Monitoring drug dose response in a large scale screen, such as control charting on drug potency (Quantitative Biology).
  • Choice of sensitizer control (positive control) which may be targets related to drug MOA.
  • Inclusion of extra control plates (see Appendix) along with other library plates in the screen to assess the quality of the screen especially HTS.

Hit Selection in Synthetic Lethality Screens

Normalization methods basically are the same as described earlier for LOF screens. The basic idea of synthetic lethality experiments is to identify hits that result in maximum chemosensitization. Therefore, we suggest the following hit selection process:


When using a cell growth or death assay, we suggest excluding si/shRNA hits that are result in high cytotoxicity without drug. This is to prevent confounding interpretation around drug potentiation (These hits can be tested separately for any sensitization effect). As an example, in our experience, we have excluded hits that cause >60% loss of viability from the following analysis. Other threshold values can also be obtained by using population-based methods suggested by statisticians (such as 2 or 3 standard deviations from the mean).


After the first step,

  • In a single-concentration experiment with sufficient replicates, one can use statistical models (such as linear models) to pick statistically significant hits that demonstrate significant interaction of drug and siRNA. Furthermore, to rank hits, we suggest a non-parametric metric based on the interaction between RNAi, drug and the combination, called “potentiation score”, based on the idea of independent events, calculated as shown below for inhibition assays, such as cell viability:
    Image cbrnai-Image001.jpg
    Or, for activation assays, such as cell apoptosis:
    Image cbrnai-Image002.jpg
    "UT" here refers to the untreated condition; the “drug only” and “combination” are at the same drug dose point. P >1 indicates that the combination effect is more than the product of two individual effects. The threshold values can be determined using population-based methods. An example of hits in single-concentration experiment is illustrated in Figure 13.
  • In a multiple-concentration experiment, sigmoidal curve comparison is done between RNAi with and without small molecule. Using cell viability assay as an example, hits that demonstrate a significant left shift of dose response curves (Figure 14) would be of interest. One should first exclude those response curves above the negative controls (to avoid transfection artifacts) and then look for a decrease of IC50/EC50 values. Statistical tests like t-test between IC50/EC50 estimates, F test for two curve fittings, or information criteria can be used for testing significance (GraphPad Prism Manual). In general, we recommend hits that show at least a 2-fold EC50 shift with respect to the negative control.
  • Apart from the follow-up mentioned above we recommend confirmation of sensitization in a multiple dose format (10-point with replicates). If available, testing related compounds for specificity is suggested.
Figure 13. . A hit from single-dose synthetic lethality screen in a cell proliferation assay.

Figure 13.

A hit from single-dose synthetic lethality screen in a cell proliferation assay. The black round points are for negative controls (NS), showing not much different effect w/ or w/o drug; the off-diagonal red triangle points are for the hit, which does (more...)

Figure 14. . A hit in multiple-dose synthetic lethality siRNA screen in a cell viability assay.

Figure 14.

A hit in multiple-dose synthetic lethality siRNA screen in a cell viability assay. The decrease of IC50 value of the red dose response curve (the siRNA with the drug) compared to the black curve (negative control, NS with the drug) is observed (the dropping (more...)

Loss-of-Function Screens Using shRNA

General Considerations for shRNA-Lentivirus Infection

Many of the same consideration for siRNA screening can be applied to arrayed shRNA screening. However, optimization of shRNA-lentivirus infection for each cell line is a more involved process than siRNA. There are various parameters that should be considered when optimizing infection.


Determination of cell seeding density from performing a simple growth curve experiment


Determination of puromycin concentration by performing a 10-point dose response curve, ranging from 0.1 mg/ml to 10 mg/ml (a typical concentration ranges between 2-5 mg/ml)


Time course for puromycin treatment


Effect of protamine sulfate to cells


The amount of virus to be used for maximal infection. A detailed protocol on viral infection can be found at http://www​​.edu/genome_bio/trc/publicProtocols.html. Furthermore, infectability can be measured for each cell line using a cell count assay:

Image cbrnai-Image003.jpg

*Refers to background (killing) i.e. just cells with puromycin added

shRNA-Lentivirus Infection Protocol

Viral infection in 96-well microplate format. This step is similar for determination of viral titer.


Seed cells of interest overnight in a total volume of 50 µl of growth media


Add 40 µl of growth media with 2X of protamine sulfate (16 mg/ml) to cells. This volume is dependent of the viral supernatant added in below in Step 3.


Add 2-10 µl of viral supernatant to mix above. This volume is dependent on the viral titer. The final volume of steps 2 and 3 is 50 µl. Incubate at 37°C overnight.


Add 2X puromycin (4 mg/ml) in 100 µl of growth media and incubate for 37°C overnight.


Wash off puromycin and replace with normal growth media.


Incubate for 2-4 days depending on the assays.

Pooled shRNA Screening

Pooled shRNAs enable large-scale screens without the need for HTS infrastructure. Pooled screens are conducted by transducing cells with a soup of shRNA-containing lentiviral particles, which can comprise 1000s of unique shRNAs. Pooled screens are performed under positive or negative selection. In positive selection, a selective pressure is applied, and the identity of shRNAs in selected cells is identified. In negative selection screens, a control population of transduced cells is compared to a treated population, and shRNAs that are lost or enriched in the treated arm are identified. Pooled shRNA libraries are commercially available and corresponding protocols are provided in detail. Some general considerations include infecting cells at a low MOI (0.1 – 0.3) to ensure no more than one integrant per cell, transducing at a reasonable fold representation (e.g., 100 – 1000 fold representation for each shRNA in the pool), and maintaining adequate representation throughout all steps of the screening process. For example, harvesting genomic DNA from a number of cells that at leasts corresponds to the intended number of viral integrants. This will ensure that all shRNAs in the experiment population are represented.


shRNA-lentivirus system

shRNA can be delivered into cells either by transfection of plasmids expressing shRNA of the gene of interest or by infection of viral-packaged shRNA of the gene of interest in the form of lentiviral vectors. The following optimization of delivery of shRNA into cells is focused on the lentiviral shRNA vectors. The lentiviral library used here is created from a pLKO1 vector that carries a puromycin resistance gene and shRNA expression is driven from a human U6 promoter (5). The puromycin resistance gene has been used as a selection marker for infected cells harboring the shRNA vectors.

Cell based phenotypic RNAi assays using shRNA lentiviral vectors involves (1) viral production where shRNA vectors are packaged into lentivirus and (2) viral infection where the lentivirus harboring the shRNA vectors are transduced into the cells of interest.

Optimization of shRNA-lentivirus production

The production of shRNA-lentivirus involves the packaging of the shRNA vector into lentivirus and requires transfection of two plasmids which forms the packaging system, pCMVD8.9 (28, 29) and pHCMV-G (30). In the transfection process, the key factors to be optimized are the seeding density, transfection reagents used, concentration of plasmids and the ratio of transfection reagent to plasmids. For concentration of plasmids, the usual practice includes concentration ranging from 100 ng to 200 ng. The plasmid concentration to transfection reagent ratio to be tested usual includes 2:1, 3:1 and 3:2. Viral production can be performed using the protocol published by the Broad institute at

GFP control vector is used for optimization purpose and can be viewed briefly under the microscope to assess fluorescence. The number of infectious units in the viral supernatant calculated as IU/ml is assessed by infecting cells with generally 2 ml of virus and counting survival of cells after puromycin treatment. The viral titer determination is important to assess the amount of virus to be used in infection for cell based assay. An acceptable range for viral titer is 2 x 106 to 2 x 107. A variety of commercial kits (p24 ELISA) are now available to determine titer.

shRNA-lentivirus production protocol (96-well microplate format)


Dilute D8.9 to 9 ng/ml, vsv-g to 1 ng/ml and shRNA to 25 ng/ml.


Add 6 µl per well (150 ng) of shRNA and 5 ml each of D8.9 (45 ng) and vsv-g (5 ng) to the shRNA.


Dilute transfection reagent (e.g. Fugene 6 from Roche) in Opti-Mem to a volume of 14 ml per well, that is, 0.6 ml of reagent to 13.4 ml of Opti-Mem. The final ratio of transfection reagent:vDNA should be 3 ml:1mg.


Add diluted transfection reagent (e.g. Fugene 6 from Roche) to the plasmid mix to a final volume of 30 ml per well and incubate for 30-45 minutes at room temperature.


Transfer Fugene/DNA complex to HEK293T cells grown overnight seeded at 25000 cells per well in low antibiotic growth media. Incubate for 18 hours at 37°C.


Replace media with 170 ml of high serum growth media and incubate for further 24 hours at 37°C.


Harvest 150 ml of viral supernatant and add 170 ml of high serum growth media and incubate for another 24 hours at 37°C.


Harvest another 150 ml of viral supernatant and discard cells.


Pool viral supernatant and use for infection.

Examples of plate layout for control plates to quality control RNAi synthetic lethality screens.


The authors would like to acknowledge collaborations with Translational Genomics Research Institute (Phoenix, AZ) and Genome Institute of Singapore (Singapore) that have led to joint learnings in adopting best practices for executing RNAi screens. We would like to acknowledge Seppo Karrila (Lilly Singapore Center for Drug Discovery) for introducing the potentiation score method. We also thank Pat Solenberg and Wayne Blosser (Eli Lilly and Company) for critical review and recommendations of content.


  • RNAi - RNA Interference
  • siRNA - Small Interfering RNA
  • shRNA - Short Hairpin RNA
  • miRNA - Micro-RNA
  • dsRNA - Double-Stranded RNA
  • KD - Knock-Down
  • HTS - High Throughput Screen/Screening
  • LOF - Loss-Of-Function
  • QC - Quality Control
  • NS - Non-Silencing
  • NT - Non-Transfected
  • NTC - Non-Targeting Control
  • MT - Mock-Transfected
  • NC - Negative Control
  • PC - Positive Control
  • CV - Coefficient of Variation
  • DDTR - Drug Dose Time Response
  • UT/T - (Drug) Untreated/Treated



Kettling RF. The many faces of RNAi. Developmental Cell. 2011;20(2):148–161. [PubMed: 21316584]
Mohr SE, Perrimon N. RNAi screening: new approaches, understandings, and organisms. WIREs RNA. 2012;3:145–158. [PMC free article: PMC3249004] [PubMed: 21953743]
Panda D, Cherry S. Cell-based genomic screening: elucidating virus-host interactions. Current Opinion in Virology. 2012;2:784–792. [PMC free article: PMC3512112] [PubMed: 23122855]
Chatterjee-Kishore M. From genome to phenome--RNAi library screening and hit characterization using signaling pathway analysis. Current Opinion in Drug Discovery & Development. 2006;9(2):231–239. [PubMed: 16566293]
Echeverri CJ, Beachy PA, Baum B, Boutros M, Buchholz F, Chanda SK, Downward J, Ellenberg J, Fraser AG, Hacohen N, Hahn WC, Jackson AL, Kiger A, Linsley PS, Lum L, Ma Y, Mathey-Prevot B, Root DE, Sabatini DM, Taipale J, Perrimon N, Bernards R. Minimizing the risk of reporting false positives in large-scale RNAi screens. Nature Methods. 2006;3(10):777–779. [PubMed: 16990807]
Echeverri CJ, Perrimon N. High-throughput RNAi screening in cultured cells: a user's guide. Nature Review Genetics. 2006;7(5):373–384. [PubMed: 16607398]
Fuchs F, Boutros M. Cellular phenotyping by RNAi. Briefings in Functional Genomics and Proteomics. 2006;5(1):52–56. [PubMed: 16769679]
Moffat J, Sabatini DM. Building mammalian signaling pathways with RNAi screens. Nature Reviews Molecular Cell Biology. 2006;7(3):177–187. [PubMed: 16496020]
Root DE, Hacohen N, Hahn WC, Lander ES, Sabatini DM. Genome-scale loss-of-function screening with a lentiviral RNAi library. Nature Methods. 2006;3(9):715–719. [PubMed: 16929317]
Sachse C, Krausz E, Kronke A, Hannus M, Walsh A, Grabner A, Ovcharenko D, Dorris D, Trudel C, Sonnichsen B, Echeverri CJ. High-throughput RNA interference strategies for target discovery and validation by using synthetic short interfering RNAs: functional genomics investigations of biological pathways. Methods in Enzymology. 2005;392:242–277. [PubMed: 15644186]
Willingham AT, Deveraux QL, Hampton GM, Aza-Blanc P. RNAi and HTS: exploring cancer by systematic loss-of-function. Oncogene. 2004;23(51):8392–8400. [PubMed: 15517021]
Birmingham A, Anderson EM, Reynolds A, Ilsley-Tyree D, Leake D, et al. 3’UTR seed matches, but not overall identity, are associated with RNAi off-targets. Nature Methods. 2006;3:199–204. [PubMed: 16489337]
Jackson AL, Burchard J, Leake D, Reynolds A, Schelter J, et al. Position-specific chemical modification of siRNAs reduces “off-target” transcript silencing. RNA. 2006;12:1197–11205. [PMC free article: PMC1484422] [PubMed: 16682562]
Jackson AL, Bartz SR, Schelter J, Kobayashi SV, Burchard J, et al. (2003). Expression profiling reveals off-target gene regulation by RNAi. Nature Biotechnology 21:635–3.7. [PubMed: 12754523]
Semizarov D, Frost L, Sarthy A, Kroeger P, Halbert DN, Fesik SW. Specificity of short interfering RNA determined through gene expression signatures. PNAS. 2003;100:6347–52. [PMC free article: PMC164449] [PubMed: 12746500]
Lin X, Ruan X, Anderson MG, McDowell JA, Kroeger PE, et al. siRNA-mediated off-target gene silencing triggered by a 7 nt complementation. Nucleic Acids Research. 2005;33:4527–4535. [PMC free article: PMC1184219] [PubMed: 16091630]
Schultz N, Marenstein DR, DeAngelis DS, Wang WQ, Nelander S, Jacobsen A, Marks DS, Massaque J, and Sander C. (2011). Off-target effects dominate a large-scale RNAi screen for modulators of the TGF-β pathway and reveal microRNA regulation of TGFBR2. Silence 2: 1758-907X-2-3. [PMC free article: PMC3068080] [PubMed: 21401928]
Sigoillot FD, King RW. Vigilance and validation: Keys to success in RNAi screening. ACS Chemical Biology. 2011;6:47–60. [PMC free article: PMC3306249] [PubMed: 21142076]
Busman FD, Malani N, Fernandes J, D’Orso I, Cagney G, Diamond TL, et al. Host cell factors in HIV replication: Meta-Analysis of genome-wide studies. PLoS Pathogens. 2009;5:e1000437. [PMC free article: PMC2682202] [PubMed: 19478882]
Marine S, Bahl A, Ferrer M, Buehler E. Common seed analysis to identify off-target effects in siRNA screens. Journal of Biomolecular Screening. 2012;17:370–378. [PubMed: 22086724]
Konig R, Chiang C, Tu BP, Yan SF, DeJesus PD, Romero A, Berguer T, et al. A probability-based approach for the analysis of large-scale RNAi screens. Nature Methods. 2007;4:847–849. [PubMed: 17828270]
Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989;45:255–268. [PubMed: 2720055]
Lin LI. A note on the concordance correlation coefficient. Biometrics. 2000;56:324–325.
Bland JM, Altaman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;I:307–310. [PubMed: 2868172]
Buehler E, Khan AA, Marine S, Rajaram M, Bahl A, Burchard J, Ferrer M. siRNA off-target effects in genome-wide screens identify signaling pathway members. Scientific Reports. 2012;2:428. article. [PMC free article: PMC3361704] [PubMed: 22645644]
Sigoillot FD, Lyman S, Huckins JF, Adamson B, Chung E, Quattrochi B, King RW. A bioinformatics method identifies prominent off-targeted transcripts in RNAi screens. Nature Methods. 2012;19:363–366. [PMC free article: PMC3482495] [PubMed: 22343343]
Buehler E, Chen YC, Martin SE. C911: A bench-level control for sequence specific siRNA off-target effects. PLoS One. 2012;7:e51942. [PMC free article: PMC3522603] [PubMed: 23251657]
Naldini L, Blömer U, Gage FH, Trono D, Verma IM. Efficient transfer, integration, and sustained long-term expression of the transgene in adult rat brains injected with a lentiviral vector. PNAS. 1996;93:11382–11388. [PMC free article: PMC38066] [PubMed: 8876144]
Zufferey R, Nagy D, Mandel RJ, Naldini L, Trono D. Multiply attenuated lentiviral vector achieves efficient gene delivery in vivo. Nature Biotechnology. 1997;15:871–875. [PubMed: 9306402]
Yee J, Miyanohara A, LaPorte P, Bouic K, Burns JC, Friedmann T. A General Method for the Generation of High-Titer, Pantropic Retroviral Vectors: Highly Efficient Infection of Primary Hepatocytes. PNAS. 1994;91:9564–9568. [PMC free article: PMC44853] [PubMed: 7937806]

Additional References:

  1. Berns K, Horlings HM, Hennessy BT, Madiredjo M, Hijmans EM, Beelen K, Linn SC, Gonzalez-Angulo AM, Stemke-Hale K, Hauptmann M, Beijersbergen RL, Mills GB, van de Vijver MJ, Bernards R. A functional genetic approach identifies the PI3K pathway as a major determinant of trastuzumab resistance in breast cancer. Cancer Cell. 2007 Oct;12(4):395–402. [PubMed: 17936563]
  2. Brideau C., Gunter B., Pikounis B., Liaw A. Improved statistical methods for hit selection in high-throughput screening. Journal of Biomolecular Screening. 2003;8:634–647. [PubMed: 14711389]
  3. Brummelkamp TR, Fabius AW, Mullenders J, Madiredjo M, Velds A, Kerkhoven RM, Bernards R, Beijersbergen RL. An shRNA barcode screen provides insight into cancer cell vulnerability to MDM2 inhibitors. Nat Chem Biol. 2006 Apr;2(4):202–6. Epub 2006 Feb 13. [PubMed: 16474381]
  4. Eggert US, Kiger AA, Richter C, Perlman ZE, Perrimon N, Mitchison TJ, Field CM. Parallel chemical genetic and genome-wide RNAi screens identify cytokinesis inhibitors and targets. PLoS Biol. 2004 Dec;2(12):e379. Epub 2004 Oct 5. [PMC free article: PMC528723] [PubMed: 15547975]
  5. Epping MT, Wang L, Plumb JA, Lieb M, Gronemeyer H, Brown R, Bernards R. A functional genetic screen identifies retinoic acid signaling as a target of histone deacetylase inhibitors. Proc Natl Acad Sci U S A. 2007 Nov 6;104(45):17777–82. Epub 2007 Oct 29. [PMC free article: PMC2077016] [PubMed: 17968018]
  6. Gunter B., Brideau C., Pikounis B., Liaw A. Statistical and graphical methods for quality control determination of high-throughput screening data. Journal of Biomolecular Screening. 2003;8:624–633. [PubMed: 14711388]
  7. Iversen PW., Eastwood BJ., Sittampalam S., Cox KL. A comparison of assay performance measures in screening assays: signal window, Z' factor and assay variability ratio. Journal of Biomolecular Screening. 2006;11:1–6. [PubMed: 16490779]
  8. Malo N., Hanley JA., Cerquozzi S., Pelletier J., Naddon R. Statistical practice in high-throughput screening data analysis. Nature Biotechnology. 2006;24:167–175. [PubMed: 16465162]
  9. Root DE, Hacohen N, Hahn WC, Lander ES, Sabatini DM. Genome-scale loss-of-function screening with a lentiviral RNAi library. Nature Methods. 2006;3(9):715–719. [PubMed: 16929317]
  10. Sachse C, Krausz E, Kronke A, Hannus M, Walsh A, Grabner A, Ovcharenko D, Dorris D, Trudel C, Sonnichsen B, Echeverri CJ. High-throughput RNA interference strategies for target discovery and validation by using synthetic short interfering RNAs: functional genomics investigations of biological pathways. Methods in Enzymology. 2005;392:242–277. [PubMed: 15644186]
  11. Whitehurst AW, Bodemann BO, Cardenas J, Ferguson D, Girard L, Peyton M, Minna JD, Michnoff C, Hao W, Roth MG, Xie XJ, White MA. Synthetic lethal screen identification of chemosensitizer loci in cancer cells. Nature. 2007 Apr 12;446(7137):815–9. [PubMed: 17429401]
  12. Willingham AT, Deveraux QL, Hampton GM, Aza-Blanc P. RNAi and HTS: exploring cancer by systematic loss-of-function. Oncogene. 2004;23(51):8392–8400. [PubMed: 15517021]

Web sites:

1. Minimum Information About an RNAi Experiment (MIARE) -

2. GraphPad Prism Manual -

Copyright Notice

All Assay Guidance Manual content, except where otherwise noted, is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported license (CC BY-NC-SA 3.0), which permits copying, distribution, transmission, and adaptation of the work, provided the original work is properly cited and not used for commercial purposes. Any altered, transformed, or adapted form of the work may only be distributed under the same or similar license to this one.

Bookshelf ID: NBK91998PMID: 22553869


Assay Guidance Manual Links

Related information

  • PMC
    PubMed Central citations
  • PubMed
    Links to PubMed

Similar articles in PubMed

See reviews...See all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...