On the art of identifying effective and
Yi Pei & Thomas Tuschl
Small interfering RNAs (siRNAs) have been widely exploited for sequence-specific gene
knockdown, predominantly to investigate gene function in cultured vertebrate cells,
and also hold promise as therapeutic agents. Because not all siRNAs that are cognate
to a given target mRNA are equally effective, computational tools have been developed
based on experimental data to increase the likelihood of selecting effective siRNAs.
Furthermore, because target-complementary siRNAs can also target other mRNAs
containing sequence segments that are partially complementary to the siRNA, most
computational tools include ways to reduce potential off-target effects in the siRNA
selection process. Though these methods facilitate selection of functional siRNAs, they
do not yet alleviate the need for experimental validation. This perspective provides a
practical guide based on current wisdom for selecting siRNAs.
The evolutionarily conserved processes whereby small
double-stranded (ds)RNAs of distinct size and structure
sequence-specifically suppress the expression of their
target genes are referred to as RNA silencing or RNA
interference (RNAi)1. Among the repertoire of known
small RNAs, siRNAs mediate gene-specific silencing pri-
marily via recognizing and inducing degradation of the
mRNAs of targeted genes. Consequently, siRNAs have
become one of the most valuable reagents to function-
ally annotate genomes and possess great potential as
Shortly after the discovery that siRNA duplexes can
specifically silence mammalian genes, it was thought that
almost any target-complementary siRNA effectively and
specifically silences its cognate target gene5. In practice,
however, different siRNAs often manifest a spectrum of
potency, and only a fraction of them are highly effective6.
Small positional shifts along the target mRNA were suffi-
cient to alter siRNA function in an apparently unpredict-
able manner6–8. Moreover, siRNAs may nonspecifically
target unrelated genes with only partial sequence-com-
plementarity (off-target effects)9–13. Hence, it is critical
to identify effective and specific siRNA sequences to per-
form reliable gene-knockdown experiments.
Initially, empirical rules had been proposed for
siRNA selection, some of which were based on the
first identified functional siRNAs5. The evolving
understanding of the RNAi mechanism, together with
statistical analyses of libraries of siRNAs with experi-
mentally determined efficiency, led to computer-
based approaches that increased the likelihood of
identifying effective and specific siRNAs6,14,15. These
tools, however, are not perfect. (i) Not every selected
siRNA meets the desired thresholds of potency and
specificity, so that experimental proof of downregu-
lation of targeted mRNA or protein remains impor-
tant, not even considering the evaluation of potential
off-target effects. (ii) A substantial fraction of active
siRNAs may be dismissed because the weighing of fac-
tors influencing activity is complex and partly unde-
fined6,9,16. Not surprisingly, experimental approaches
to generate and identify effective siRNAs have been
developed to complement rule-driven siRNA selec-
There are many excellent recent reviews covering the
mechanism of RNAi19–22. Elements of this mechanism
that are important for the selection of siRNA are sum-
marized in Box 1.
Howard Hughes Medical Institute, Laboratory of RNA Molecular Biology, The Rockefeller University, 1230 York Avenue, Box 186, New York,
New York 10021, USA. Correspondence should be addressed to T.T. (email@example.com).
PUBLISHED ONLINE 23 AUGUST 2006; DOI:10.1038/NMETH911
670 | VOL.3 NO.9 | SEPTEMBER 2006 | NATURE METHODS
© 2006 Nature Publishing Group http://www.nature.com/naturemethods
Here we provide a practical guide and an overview of the theo-
retical basis for identification and selection of effective and specific
A GUIDE FOR siRNA SELECTION
Target mRNA analysis
The selection of siRNAs against a gene of interest starts with an
annotated target mRNA sequence, including its 5′ and 3′ untrans-
lated regions (UTRs), splice, polymorphic and allelic variants.
Because the coding sequence is the most reliable mRNA sequence
information available, it is commonly targeted. The UTRs are gen-
erally less well characterized, but can also be targeted with similar
gene-knockdown efficiency8,23,24. Though it has often been recom-
mended to avoid targeting sequences that contain known binding
sites for mRNA-binding proteins, such as the exon-exon junction
complex, there is no detailed experimental study available to assess
the importance of this guideline.
For practical reasons, selection of siRNAs is often carried out with
additional constraints, for example identifying siRNAs that target (i)
orthologs in more than one species or (ii) all possible splice variants
of a gene.
Database search for published and validated siRNAs
Several databases archive experimentally tested siRNA sequences
from the literature25–27. Additionally, validated siRNAs can be
acquired from commercial resources (for example, the Silencer vali-
dated siRNAs from Ambion and HP validated siRNAs from Qiagen).
Some vendors, such as Ambion, Qiagen and Dharmacon also pro-
vide predesigned siRNAs or custom siRNA design service. Though
prevalidated reagents provide an excellent starting point, the user
still has to examine whether these siRNAs are potent and specific to
meet the needs28.
If there are no matches to the target gene of interest in any of these
databases or in the literature, it is advisable to select 3–5 candidate
siRNAs using available guidelines and tools, and subsequently to
validate the reagents.
Selected algorithms and siRNA sequence selection tools
Several siRNA sequence selection algorithms have been developed in
recent years that rely on intrinsic sequence and stability features of
functional siRNAs6,14,15,23,29–35. A smaller number of algorithms
consider the secondary structure and accessibility of the targeted
mRNA36–38. The approaches underlying these algorithms range
from empirical observations to sophisticated machine learning.
After the siRNA sequence selection from the target mRNA sequence,
each candidate siRNA is examined for similarity to all other mRNA
transcripts that might unintentionally be targeted at a genome-wide
level. Most of the siRNA selection algorithms have been combined
with a variant of such programs, and the more user-friendly tools are
listed in Table 1 (for a more complete list, see ref. 28). The selected
siRNAs can be custom synthesized from four siRNA-licensed reagent
suppliers: Ambion, Dharmacon, Qiagen and Sigma Proligo.
Prevalidation of siRNAs
Because the determination of the precise level of gene knock-
down for each siRNA is a demanding process, and the assays need
to be adapted for newly targeted genes, reporter-based assays
have been developed to accelerate the identification of potent
siRNAs among various synthesized siRNAs. In these systems,
plasmids, which carry the target sequence fused to a reporter
RNAi is a gene-regulatory mechanism triggered by dsRNAs.
siRNAs, which consist of duplexes of 21–23 nt RNAs that are
base-paired with 2-nt 3′ overhangs, mimic intermediates of
the natural processing of longer double-stranded RNA triggers
by RNase III. The major products of RNase III processing are
microRNAs (miRNAs), which are endogenous ~ 22-nt RNAs that
repress gene expression by targeting mRNAs for cleavage or
An siRNA is generally designed to be fully complementary
to its target mRNA and is commonly a product of chemical
synthesis. Though the natural processing intermediates carry
5′ phosphates, the 5′ phosphate is typically omitted in chemical
synthesis as a cellular kinase rapidly phosphorylates the siRNA
once it is delivered to the cells86. Individual siRNAs or random
siRNA pools can also be generated by enzymatic methods from
digestion of longer dsRNAs65,87,88. Alternatively, siRNAs can be
generated by processing of ectopically expressed short hairpin
Naturally occurring siRNAs cognate to cellular or viral mRNAs
have not been experimentally detected in mammals. Instead, the
mammalian RNAi machinery appears to have been adapted solely
for miRNA-mediated regulation of mRNAs containing miRNA
binding sites predominantly located in the 3′ UTR92. Although
the majority of miRNAs affect the stability and translation of
mRNAs that are only partially complementary72,92, some miRNAs
use, like siRNAs, near-perfect complementarity to cleave their
targets93. The latter requirement might explain the evolutionarily
conserved catalytic aspect of targeted mRNA degradation in
RNase III and/or other components of the RNAi machinery
specifically recognize an siRNA duplex and selectively incorporate
one of the siRNA strands into different RISCs, including the
catalytic endonuclease-containing complex, which is responsible
for the strong siRNA gene-knockdown effect19,20. The strand
antisense to the targeted mRNA is often referred to as the guide
strand, and its base-paired sense strand is known as the passenger
strand, which is destroyed upon incorporation of the guide strand
into RISC47,94,95 (Fig. 1). The catalytic RISC recognizes mRNAs
containing perfect or near-perfect complementary sequence to
the guide siRNA and cleaves the mRNAs at a site precisely 10 nt
upstream of the nucleotide opposite the 5′-most nucleotide of
the guide strand (Fig. 1). The mRNA fragments are subsequently
degraded by cellular nucleases, resulting in knockdown of the
expression of the corresponding genes.
BOX 1 siRNA-MEDIATED GENE SILENCING IN MAMMALIAN CELLS: ELEMENTS OF
NATURE METHODS | VOL.3 NO.9 | SEPTEMBER 2006 | 671
© 2006 Nature Publishing Group http://www.nature.com/naturemethods
gene and a control gene for normalization, are cointroduced
into cells together with the target-specific siRNAs14,39,40. The
dual-luciferase–based siCHECK system from Promega is widely
used and provides a ranking for siRNA activity within 24 h. The
reporter-based activity generally correlates well with the efficacy
of depleting the endogenous target (our unpublished observa-
tions). The prevalidated siRNAs can then be used to validate the
depletion of the endogenous target mRNA, which is discussed in
detail in an accompanying review41.
CONSIDERATIONS FOR SELECTING EFFECTIVE AND
Sequence asymmetry of siRNA duplexes
It has been demonstrated that structurally symmetric (duplexes
with symmetric 2-nucleotide (nt) 3′ overhangs) but primary
sequence–asymmetric (different nucleotides on each end) siRNAs,
from which the target-mRNA complementary guide strand has
greater propensity to be assembled into the RNA-induced silenc-
ing complex (RISC) than the passenger strand, show improved effi-
cacy and specificity42,43 (Fig. 1). The same finding emerges from
sequence analysis of miRNA precursors and largely explains the
asymmetric accumulation of the majority of miRNAs42. The asym-
metry is determined by the different sequence composition, and the
consequent differences in thermodynamic stability and molecular
dynamic behavior of the two base-paired ends of an siRNA duplex:
the strand with the less stable 5′ end, owing to either weaker base-
pairing or introduction of mismatches, is favorably or exclusively
loaded into RISC44. The asymmetry rule has been implemented in
many siRNA design algorithms by computing either the A⋅U base
pair content or local free energy at both ends of an siRNA, followed
by selection of the duplexes with less stable, (A+U)-enriched 5′ end
on the guide strand20.
Because the majority of miRNAs start with a 5′ uridine, it is also
conceivable that 5′ uridine–specific interaction contributes to more
effective RISC assembly and function beyond the thermodynamic
contributions discussed here. Furthermore, miRNA duplexes con-
tain an average of six non–Watson-Crick base pairs distributed over
the entire miRNA length, whose contribution to RISC assembly and
asymmetry has not been evaluated.
siRNA duplex stability
Most analyzed functional siRNAs had a low-to-medium G+C
content ranging between 30% and 52% (refs. 6,31). It has been
argued that too low G+C content may destabilize siRNA duplexes
and reduce the affinity for target mRNA binding, whereas too high
G+C content may impede RISC loading and/or cleavage-product
release. Additionally, surveys of functional siRNAs revealed that sta-
ble duplexes devoid of internal repeats or palindromes, which may
form intrastrand secondary structures, were better silencers6,31,45.
An equally likely explanation is that the secondary structure of the
target mRNA, which mirrors the predicted guide siRNA secondary
structure, interferes with targeting.
Although the overall duplex stability is important, the center of
the duplex (positions 9–14 on the guide strand) appears to prefer-
entially have low internal stability31,42,46. It has recently been noticed
that miRNAs and siRNAs assemble into RISC by different mecha-
nisms; siRNAs require cleavage of the passenger strand for effective
RISC assembly, whereas a mismatched RNase III–processed miRNA
duplex does not require passenger strand cleavage47. It is conceiv-
able that the central-duplex instability may influence how effectively
and to what ratios the RISC complexes with different core compo-
nents are loaded. Alteration of the structure and stability of siRNA
duplexes can also be controlled by incorporation of chemically
modified nucleotide analogs. The effects of modifications, however,
Table 1 | Representative siRNA sequence selection web tools
Scores and ranks candidate siRNAs based on thermodynamic and
sequence-related criteria. BLAST search is conducted by default.
Ranks candidate siRNAs using a primitive scoring system.
BLAST search is automatic and the results are shown.
An artificial neural network–based tool, which was trained with
~2,500 experimentally assessed siRNAs. Analysis of genome-wide
specificity is included.
Offers flexibility in defining siRNA sequence patterns and
selection of filter functions. Different properties of selected
siRNAs are calculated, including thermodynamic values,
polymorphisms are identified and the results of configurable
BLAST search and filtering are shown. The user can sort the
output in various ways and balance decisions.
Developed for high-throughput applications of siRNAs using
several published algorithms for efficacy prediction and a
nonredundant database for specificity analysis.
The kernel algorithm focuses primarily on energy features of
effective siRNAs. Alternative algorithms are also implemented
and integrated in the tool. siSearch is expandable to include
newly discovered rules.
Sequence selection tool, which incorporates the target
accessibility in the evaluation. No specificity analysis.
Candidate siRNAs proposed by various previously developed
sequence selection tools are classified based on target
RNAi Designer https://rnaidesigner.invitrogen.com/
Whitehead siRNA Selection serverhttp://jura.wi.mit.edu/bioc/siRNA 30
Sirna http://sfold.wadsworth.org/sirna.pl 36
siRNA design softwarehttp://www.cs.hku.hk/~sirna 38
672 | VOL.3 NO.9 | SEPTEMBER 2006 | NATURE METHODS
© 2006 Nature Publishing Group http://www.nature.com/naturemethods
are dependent on the position and the sequence context, and general
rules are not yet available2,8,48.
It is interesting to mention that imperfectly paired siRNA duplexes
composed of target mRNA–complementary and partially palin-
dromic or partially complementary single-stranded siRNAs have
also been used successfully49. These siRNA duplexes are solely com-
posed of two fully target-complementary guide strands that are
sufficiently complementary to each other to form stable duplexes
with characteristic 3′ overhanging ends. The silencing efficiencies
of guide-only siRNA duplexes are comparable to prototypical fully
paired passenger-guide duplex siRNAs, even though guide-only
siRNA duplexes may contain a substantial number of non–Watson-
Crick and G⋅U wobble base pairs.
It has been argued that local secondary structures (short stem-loops)
in target mRNAs might restrict the accessibility of RISC, and attenu-
ate or abolish siRNA efficacy37,50–53. A major obstacle in assessing
target-accessibility is the lack of tools that reliably predict mRNA sec-
ondary structure, setting aside the fact that mRNA is present inside
cells as ribonucleoprotein complex of unknown composition. Several
algorithms have been developed, and filtering of potentially inac-
cessible target sequences has been shown to
improve functional siRNA selection36–38,51.
Several sequence analyses of siRNAs have
independently identified single nucleo-
tide positional preferences, which we will
summarize using the guide strand as refer-
ence6,14,23,29,32–35 (Fig. 2): (i) U or A at posi-
tion 1; (ii) C or G (C is more common) at
position 19; (iii) A+U richness between posi-
tions 1 and 7; (iv) A or U (A is more com-
mon) at position 10; (v) other motifs that
were overrepresented in one analysis but not
others, such as a U at position 17. The first
three sequence features correlate with the
rule of thermodynamic asymmetry, and the
preferred nucleotides on indicated positions
may contribute to the bias for selection of
antisense strand. The A or U at position 10 is at the cleavage site and
may promote catalytic RISC-mediated passenger strand and substrate
cleavage. Other sequence determinants may be involved in steps along
the RNAi pathway, such as RISC loading54.
In addition to the positional nucleotide preference, certain motifs
are commonly avoided in chemically synthesized siRNA duplexes
that could affect the synthesis yield, purification or the annealing of
siRNA strands. Extended runs of altering G⋅C pairs (more than 7)32
or runs of more than three guanines are sometimes avoided.
Moreover, in light of the reports that certain siRNAs can activate
immune response in a cell- and sequence-dependent manner55–57,
it is a prudent measure to filter out siRNA sequences containing
putative immunostimulatory motifs in either strand to minimize
toxicities and nonspecific silencing effects, especially when siRNAs
are selected for in vivo and therapeutic use. Alternatively, immuno-
stimulatory side effects can be masked using chemically modified
nucleotides56,58,59. It is uncertain if all immunostimulatory RNA
motifs have yet been defined.
siRNA length and asymmetric 3′ overhang
Conventionally designed siRNAs are 21-mers with symmetric 2-nt
3′ overhangs, representing the predominant processing intermediate
in the RNAi pathway60,61. It was noticed early that 20–25 nt siRNA
duplexes carrying 2-nt 3′ overhangs could reach similar gene silenc-
ing efficiency in mammalian cell culture experiments5,62, and that
expressed or synthetic hairpin RNAs of up to 29 base pairs in length
also triggered effective gene silencing63. These observations were fol-
lowed up more recently, testing synthetic 21–29 nt RNA duplexes
with blunt ends, symmetric or asymmetric 2-nt overhangs. These
siRNA precursor molecules can silence target genes with similar
efficiency to conventional siRNA duplexes64–68. The longer dsRNAs
appear to have the advantage that they can be transfected at lower
concentrations than conventional siRNAs without loss of gene
silencing. But they also appear to be more likely to induce nonspe-
cific responses (including interferon induction) or mediate other
effects on cell viability67,68.
Short RNA duplexes composed of single-strands of different
length (19/21, 21/23 or 25/27 nt formats) have also been shown to
silence mRNAs effectively66,69,70. The single 2-nt 3′ overhang present
in those duplexes at the 3′ end of the guide strand and its presumed
Figure 1 | A scheme for siRNA-mediated gene silencing. The primary
sequence asymmetry of duplex determines which strand is preferentially
assembled into RISC.
Figure 2 | siRNA and target mRNA structures. (a) Standard siRNA duplex. (b) Target mRNA specificity.
The cleavage site is indicated by scissors in the target mRNA. Target recognition and off-target
activity can occur in two modes, the catalytic siRNA-guided cleavage reaction requiring extensive
complementarity in the region surrounding the cleavage site (blue) and the miRNA-like destabilization
of mRNAs requiring pairing of the siRNA 5′ end (green).
Guide strand (preferentially enters RISC)
Passenger strand destroyed
Passenger strand (preferentially destroyed)
3′ OH P 5′
21 20 19
18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
NATURE METHODS | VOL.3 NO.9 | SEPTEMBER 2006 | 673
© 2006 Nature Publishing Group http://www.nature.com/naturemethods
interaction with the RNAi machinery may contribute to asymmet-
ric RISC assembly. However, the comparison with conventional
siRNAs is again complicated by the differences in length of the guide
strand and the differences in strength of 5′ terminal guide strand
base-pairing when the paired region is longer than in conventional
Each strand of an siRNA duplex, once assembled into RISC, can
guide recognition of fully and partially complementary target
mRNAs, referred to as on- and off-targets, respectively. Though
sequence asymmetry can be used to bias passenger strand exclu-
sion, chemical methods of preventing passenger-strand use have also
been introduced (for example, Dharmacon’s ON-TARGET siRNA).
For the purpose of this discussion, we will distinguish off-targets
into two classes (Fig. 2): (i) those that share contiguous and centrally
located sequence complementarity over more than half of the siRNA
sequence somewhere within the mRNA sequence71, and (ii) those
that show solely 6 or 7 nucleotides of perfect match preferentially
in the 3′ UTRs with positions 2–7 or 2–8 (seed region) of the guide
siRNA9,11,12. The latter interaction is the major driving force behind
endogenous miRNA–target mRNA recognition20,72. Although the
off-targets of the latter class are predominant, their actual number
identified in microarray analyses was significantly smaller than the
number of computationally predicted targets with sequence com-
plementary to the seed region of the guide strand, suggesting that
additional specificity determinants remain to be identified9,12.
Furthermore, structural and biochemical studies showed that
guide-strand position 1 and the nucleotides at the 3′ overhang (posi-
tions 20 and 21) have little, if any, contribution to the specificity of
target recognition, and that mismatches near the 5′ and 3′ ends can
be tolerated for RISC-guided cleavage if the remaining pairing to the
target was unperturbed73,74.
To enforce specificity, the current strategy is to select siRNAs in
which the strand(s) entering RISC has some mismatches to all unde-
sired target mRNAs, especially their 3′ UTRs. Typically at least three
mismatches are recommended between positions 2 and 19 and the
mismatches near the 5′ end and in the center of the examined strand
should be assigned higher significance11,71,75,76. In addition to the
position, the identity of the sequence mismatches also influence
specificity to a certain extent75,77,78.
Presently most tools use blastn or Smith-Waterman algo-
rithm to remove potential off-targeting siRNAs during the siRNA
sequence selection process79. In addition to the search method,
the quality and completeness of the selected genome-wide mRNA
sequence database is also of high importance79–81. The current
tools, however, cannot eliminate all the potential off-targets, espe-
cially those that contain the short sequence segments comple-
mentary to the seed region of the guide strand, and likely discard
many potentially functional siRNAs9. While improved algorithms
are awaited, position-specific chemical modification of the
seed-sequence of the guide siRNA can be used to reduce off-tar-
get effects82. It is therefore important to experimentally control
off-target effects or to dilute the off-target effects beyond the detection
limit by codelivering several different target-specific siRNAs10,41.
Allele-specific gene silencing
To take advantage of the sequence specificity of RNAi, a prerequi-
site to achieve allele-specific gene silencing is to identify the most
significant difference between two alleles, which may be as little
as a single nucleotide change stemming from mutation or poly-
morphism83. Placing this sequence discrepancy in the center of an
siRNA, at or near the RISC cleavage site seems to be best for discrim-
inating between alleles8,76–78,83. In some cases, introducing an addi-
tional mismatch at other positions in the siRNA may improve the
allele specificity, as long as the mismatch is tolerated for cleavage83.
A limitation of this approach is that the choice of siRNA is restricted,
and the siRNA may not be sufficiently effective. In this respect, it is
interesting to note that the introduction of a G⋅U wobble mismatch
in the 5′ terminal siRNA-mRNA interaction increased the potency
of some siRNAs75. The efficacy of silencing may also be increased by
destabilizing base-pairing at the 5′ end of the guide strand following
the asymmetry rule78.
Alternatively, both alleles can be nondiscriminately silenced by an
effective siRNA distant from the polymorphic site, accompanied by
ectopic expression of the desired sequence-modified allele refractory
to the siRNA84. Vectors that simultaneously express transgene and
short hairpin RNAs have been developed85.
In summary, guidelines are available that increase the likelihood of
identifying effective and specific siRNAs at the expense of elimi-
nating many potentially functional and specific siRNAs. These
guidelines assist in reducing the numbers of siRNAs that need to
be experimentally validated to identify potent and specific siRNAs
for a given target gene. As reagent manufacturers have recognized
the need for constant validation of siRNA knockdown experiments
and developed promising lines of reagents, effective siRNAs can be
identified at a rapid pace and will soon lead to the ultimate goal of
production of validated genome-wide siRNA libraries needed for
high-throughput or individual gene silencing experiments.
We apologize to authors whose works are not cited owing to space limitations.
We thank C. Echeverri at Cenix Bioscience for valuable discussion. We also thank
M. Landthaler, P. Landgraf, J. Brennecke and C. Rogler for critical reading of the
manuscript. Y.P. is supported by the Ruth L. Kirschstein Fellowship from the US
National Institutes of Health–National Institute of General Medical Sciences.
COMPETING INTERESTS STATEMENT
The authors declare competing financial interests (see the Nature Methods
website for details).
Published online at http://www.nature.com/naturemethods/
Reprints and permissions information is available online at http://npg.
1. Meister, G. & Tuschl, T. Mechanisms of gene silencing by double-stranded
RNA. Nature 431, 343–349 (2004).
2. Dorsett, Y. & Tuschl, T. siRNAs: applications in functional genomics and
potential as therapeutics. Nat. Rev. Drug Discov. 3, 318–329 (2004).
3. Echeverri, C.J. & Perrimon, N. High-throughput RNAi screening in cultured
cells: a user’s guide. Nat. Rev. Genet. 7, 373–384 (2006).
4. Dykxhoorn, D.M., Palliser, D. & Lieberman, J. The silent treatment: siRNAs as
small molecule drugs. Gene Ther. 13, 541–552 (2006).
5. Elbashir, S.M., Harborth, J., Weber, K. & Tuschl, T. Analysis of gene function
in somatic mammalian cells using small interfering RNAs. Methods 26, 199–
6. Reynolds, A. et al. Rational siRNA design for RNA interference. Nat.
Biotechnol. 22, 326–330 (2004).
7. Holen, T., Amarzguioui, M., Wiiger, M.T., Babaie, E. & Prydz, H. Positional
effects of short interfering RNAs targeting the human coagulation trigger
Tissue Factor. Nucleic Acids Res. 30, 1757–1766 (2002).
8. Harborth, J. et al. Sequence, chemical, and structural variation of small
interfering RNAs and short hairpin RNAs and the effect on mammalian gene
674 | VOL.3 NO.9 | SEPTEMBER 2006 | NATURE METHODS
© 2006 Nature Publishing Group http://www.nature.com/naturemethods
silencing. Antisense Nucleic Acid Drug Dev. 13, 83–105 (2003).
9. Birmingham, A. et al. 3′ UTR seed matches, but not overall identity, are
associated with RNAi off-targets. Nat. Methods 3, 199–204 (2006).
10. Jackson, A.L. & Linsley, P.S. Noise amidst the silence: off-target effects of
siRNAs? Trends Genet. 20, 521–524 (2004).
11. Lin, X. et al. siRNA-mediated off-target gene silencing triggered by a 7 nt
complementation. Nucleic Acids Res. 33, 4527–4535 (2005).
12. Jackson, A.L. et al. Widespread siRNA “off-target” transcript silencing
mediated by seed region sequence complementarity. RNA 12, 1179–1187
13. Fedorov, Y. et al. Off-target effects by siRNA can induce toxic phenotype. RNA
12, 1188–1196 (2006).
14. Huesken, D. et al. Design of a genome-wide siRNA library using an artificial
neural network. Nat. Biotechnol. 23, 995–1001 (2005).
15. Saetrom, P. & Snove, O., Jr. A comparison of siRNA efficacy predictors.
Biochem. Biophys. Res. Commun. 321, 247–253 (2004).
16. Zhao, H.F. et al. High-throughput screening of effective siRNAs from RNAi
libraries delivered via bacterial invasion. Nat. Methods 2, 967–973 (2005).
17. Ito, M., Kawano, K., Miyagishi, M. & Taira, K. Genome-wide application of
RNAi to the discovery of potential drug targets. FEBS Lett. 579, 5988–5995
18. Kasim, V., Taira, K. & Miyagishi, M. Screening of siRNA target sequences by
using fragmentized DNA. J. Gene Med. 8, 782–791 (2006).
19. Sontheimer, E.J. Assembly and function of RNA silencing complexes. Nat.
Rev. Mol. Cell Biol. 6, 127–138 (2005).
20. Tomari, Y. & Zamore, P.D. Perspective: machines for RNAi. Genes Dev. 19,
21. Filipowicz, W., Jaskiewicz, L., Kolb, F.A. & Pillai, R.S. Post-transcriptional gene
silencing by siRNAs and miRNAs. Curr. Opin. Struct. Biol. 15, 331–341 (2005).
22. Valencia-Sanchez, M.A., Liu, J., Hannon, G.J. & Parker, R. Control of
translation and mRNA degradation by miRNAs and siRNAs. Genes Dev. 20,
23. Hsieh, A.C. et al. A library of siRNA duplexes targeting the phosphoinositide
3-kinase pathway: determinants of gene silencing for use in cell-based
screens. Nucleic Acids Res. 32, 893–901 (2004).
24. McManus, M.T. et al. Small interfering RNA-mediated gene silencing in
T lymphocytes. J. Immunol. 169, 5754–5760 (2002).
25. Ren, Y. et al. siRecords: an extensive database of mammalian siRNAs with
efficacy ratings. Bioinformatics 22, 1027–1028 (2006).
26. Chalk, A.M., Warfinge, R.E., Georgii-Hemming, P. & Sonnhammer, E.L.
siRNAdb: a database of siRNA sequences. Nucleic Acids Res. 33, D131–D134
27. Truss, M. et al. HuSiDa–the human siRNA database: an open-access database
for published functional siRNA sequences and technical details of efficient
transfer into recipient cells. Nucleic Acids Res. 33, D108–D111 (2005).
28. Smith, C. Sharpening the tools of RNA interference. Nat. Methods 3, 475–486
29. Shabalina, S.A., Spiridonov, A.N. & Ogurtsov, A.Y. Computational models
with thermodynamic and composition features improve siRNA design. BMC
Bioinformatics 7, 65 (2006).
30. Yuan, B., Latek, R., Hossbach, M., Tuschl, T. & Lewitter, F. siRNA Selection
Server: an automated siRNA oligonucleotide prediction server. Nucleic Acids
Res. 32, W130–W134 (2004).
31. Chalk, A.M., Wahlestedt, C. & Sonnhammer, E.L. Improved and automated
prediction of effective siRNA. Biochem. Biophys. Res. Commun. 319, 264–274
32. Ui-Tei, K. et al. Guidelines for the selection of highly effective siRNA
sequences for mammalian and chick RNA interference. Nucleic Acids Res. 32,
33. Takasaki, S., Kotani, S. & Konagaya, A. An effective method for selecting
siRNA target sequences in mammalian cells. Cell Cycle 3, 790–795 (2004).
34. Jagla, B. et al. Sequence characteristics of functional siRNAs. RNA 11, 864–
35. Amarzguioui, M. & Prydz, H. An algorithm for selection of functional siRNA
sequences. Biochem. Biophys. Res. Commun. 316, 1050–1058 (2004).
36. Ding, Y., Chan, C.Y. & Lawrence, C.E. Sfold web server for statistical folding
and rational design of nucleic acids. Nucleic Acids Res. 32, W135–W141
37. Luo, K.Q. & Chang, D.C. The gene-silencing efficiency of siRNA is strongly
dependent on the local structure of mRNA at the targeted region. Biochem.
Biophys. Res. Commun. 318, 303–310 (2004).
38. Yiu, S.M. et al. Filtering of ineffective siRNAs and improved siRNA design
tool. Bioinformatics 21, 144–151 (2005).
39. Kumar, R., Conklin, D.S. & Mittal, V. High-throughput selection of effective
RNAi probes for gene silencing. Genome Res. 13, 2333–2340 (2003).
40. Malik, I., Garrido, M., Bahr, M., Kugler, S. & Michel, U. Comparison of test
systems for RNAinterference. Biochem. Biophys. Res. Commun. 341, 245–253
41. Cullen, B.R. Enhancing and confirming the specificity of RNAi experiments.
Nat. Methods 3, 677–681 (2006).
42. Khvorova, A., Reynolds, A. & Jayasena, S.D. Functional siRNAs and miRNAs
exhibit strand bias. Cell 115, 209–216 (2003).
43. Schwarz, D.S. et al. Asymmetry in the assembly of the RNAi enzyme complex.
Cell 115, 199–208 (2003).
44. Hutvagner, G. Small RNA asymmetry in RNAi: function in RISC assembly and
gene regulation. FEBS Lett. 579, 5850–5857 (2005).
45. Patzel, V. et al. Design of siRNAs producing unstructured guide-RNAs results in
improved RNA interference efficiency. Nat. Biotechnol. 23, 1440–1444 (2005).
46. Taxman, D.J. et al. Criteria for effective design, construction, and gene
knockdown by shRNA vectors. BMC Biotechnol. 6, 7 (2006).
47. Matranga, C., Tomari, Y., Shin, C., Bartel, D.P. & Zamore, P.D. Passenger-
strand cleavage facilitates assembly of siRNA into Ago2-containing RNAi
enzyme complexes. Cell 123, 607–620 (2005).
48. Manoharan, M. RNA interference and chemically modified small interfering
RNAs. Curr. Opin. Chem. Biol. 8, 570–579 (2004).
49. Hossbach, M., Gruber, J., Osborn, M., Weber, K. & Tuschl, T. Gene silencing
with siRNA duplexes composed of target-mRNA-complementary and partially
palindromic or partially complementary single-stranded siRNAs. RNA Biology
50. Cullen, B.R. Induction of stable RNA interference in mammalian cells. Gene
Ther. 13, 503–508 (2006).
51. Heale, B.S., Soifer, H.S., Bowers, C. & Rossi, J.J. siRNA target site secondary
structure predictions using local stable substructures. Nucleic Acids Res. 33,
52. Schubert, S., Grunweller, A., Erdmann, V.A. & Kurreck, J. Local RNA target
structure influences siRNA efficacy: systematic analysis of intentionally
designed binding regions. J. Mol. Biol. 348, 883–893 (2005).
53. Overhoff, M. et al. Local RNA target structure influences siRNA efficacy: a
systematic global analysis. J. Mol. Biol. 348, 871–881 (2005).
54. Mittal, V. Improving the efficiency of RNA interference in mammals. Nat. Rev.
Genet. 5, 355–365 (2004).
55. Judge, A.D. et al. Sequence-dependent stimulation of the mammalian innate
immune response by synthetic siRNA. Nat. Biotechnol. 23, 457–462 (2005).
56. Hornung, V. et al. Sequence-specific potent induction of IFN-α by short
interfering RNA in plasmacytoid dendritic cells through TLR7. Nat. Med. 11,
57. Marques, J.T. & Williams, B.R. Activation of the mammalian immune system
by siRNAs. Nat. Biotechnol. 23, 1399–1405 (2005).
58. Judge, A.D., Bola, G., Lee, A.C. & MacLachlan, I. Design of noninflammatory
synthetic siRNA mediating potent gene silencing in vivo. Mol. Ther. 13,
59. Snøve, O. Jr. & Rossi, J.J. Expressing short hairpin RNAs in vivo. Nat. Methods
3, 689–695 (2006).
60. Elbashir, S.M. et al. Duplexes of 21-nucleotide RNAs mediate RNA
interference in cultured mammalian cells. Nature 411, 494–498 (2001).
61. Elbashir, S.M., Lendeckel, W. & Tuschl, T. RNA interference is mediated by 21-
and 22-nucleotide RNAs. Genes Dev. 15, 188–200 (2001).
62. Caplen, N.J., Parrish, S., Imani, F., Fire, A. & Morgan, R.A. Specific inhibition
of gene expression by small double-stranded RNAs in invertebrate and
vertebrate systems. Proc. Natl. Acad. Sci. USA 98, 9742–9747 (2001).
63. Paddison, P.J., Caudy, A.A., Bernstein, E., Hannon, G.J. & Conklin, D.S. Short
hairpin RNAs (shRNAs) induce sequence-specific silencing in mammalian
cells. Genes Dev. 16, 948–958 (2002).
64. Kim, D.H. et al. Synthetic dsRNA Dicer substrates enhance RNAi potency and
efficacy. Nat. Biotechnol. 23, 222–226 (2005).
65. Amarzguioui, M., Rossi, J.J. & Kim, D. Approaches for chemically synthesized
siRNA and vector-mediated RNAi. FEBS Lett. 579, 5974–5981 (2005).
66. Rose, S.D. et al. Functional polarity is introduced by Dicer processing of short
substrate RNAs. Nucleic Acids Res. 33, 4140–4156 (2005).
67. Reynolds, A. et al. Induction of the interferon response by siRNA is cell type-
and duplex length-dependent. RNA 12, 988–993 (2006).
68. Marques, J.T. et al. A structural basis for discriminating between self and
nonself double-stranded RNAs in mammalian cells. Nat. Biotechnol. 24,
69. Vermeulen, A. et al. The contributions of dsRNA structure to Dicer specificity
and efficiency. RNA 11, 674–682 (2005).
70. Soutschek, J. et al. Therapeutic silencing of an endogenous gene by systemic
administration of modified siRNAs. Nature 432, 173–178 (2004).
71. Jackson, A.L. et al. Expression profiling reveals off-target gene regulation by
RNAi. Nat. Biotechnol. 21, 635–637 (2003).
NATURE METHODS | VOL.3 NO.9 | SEPTEMBER 2006 | 675
© 2006 Nature Publishing Group http://www.nature.com/naturemethods
72. Bartel, D.P. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell Download full-text
116, 281–297 (2004).
73. Haley, B. & Zamore, P.D. Kinetic analysis of the RNAi enzyme complex. Nat.
Struct. Mol. Biol. 11, 599–606 (2004).
74. Martinez, J. & Tuschl, T. RISC is a 5′ phosphomonoester-producing RNA
endonuclease. Genes Dev. 18, 975–980 (2004).
75. Holen, T. et al. Tolerated wobble mutations in siRNAs decrease specificity,
but can enhance activity in vivo. Nucleic Acids Res. 33, 4704–4710 (2005).
76. Elbashir, S.M., Martinez, J., Patkaniowska, A., Lendeckel, W. & Tuschl, T.
Functional anatomy of siRNAs for mediating efficient RNAi in Drosophila
melanogaster embryo lysate. EMBO J. 20, 6877–6888 (2001).
77. Du, Q., Thonberg, H., Wang, J., Wahlestedt, C. & Liang, Z. A systematic
analysis of the silencing effects of an active siRNA at all single-nucleotide
mismatched target sites. Nucleic Acids Res. 33, 1671–1677 (2005).
78. Dykxhoorn, D.M., Schlehuber, L.D., London, I.M. & Lieberman, J.
Determinants of specific RNA interference-mediated silencing of human
beta-globin alleles differing by a single nucleotide polymorphism. Proc. Natl.
Acad. Sci. USA 103, 5953–5958 (2006).
79. Boese, Q. et al. Mechanistic insights aid computational short interfering RNA
design. Methods Enzymol. 392, 73–96 (2005).
80. Santoyo, J., Vaquerizas, J.M. & Dopazo, J. Highly specific and accurate
selection of siRNAs for high-throughput functional assays. Bioinformatics 21,
81. Naito, Y., Yamada, T., Ui-Tei, K., Morishita, S. & Saigo, K. siDirect: highly
effective, target-specific siRNA design software for mammalian RNA
interference. Nucleic Acids Res. 32, W124–W129 (2004).
82. Jackson, A.L. et al. Position-specific chemical modification of siRNAs reduces
“off-target” transcript silencing. RNA 12, 1197–1205 (2006).
83. Rodriguez-Lebron, E. & Paulson, H.L. Allele-specific RNA interference for
neurological disease. Gene Ther. 13, 576–581 (2006).
84. Kubodera, T., Yokota, T., Ishikawa, K. & Mizusawa, H. New RNAi strategy
for selective suppression of a mutant allele in polyglutamine disease.
Oligonucleotides 15, 298–302 (2005).
85. Samakoglu, S. et al. A genetic strategy to treat sickle cell anemia by
coregulating globin transgene expression and RNA interference. Nat.
Biotechnol. 24, 89–94 (2006).
86. Martinez, J., Patkaniowska, A., Urlaub, H., Luhrmann, R. & Tuschl, T. Single-
stranded antisense siRNAs guide target RNA cleavage in RNAi. Cell 110,
87. Kittler, R., Heninger, A.K., Franke, K., Habermann, B. & Buchholz, F.
Production of endoribonuclease-prepared short interfering RNAs for gene
silencing in mammalian cells. Nat. Methods 2, 779–784 (2005).
88. Buchholz, F., Kittler, R., Slabicki, M. & Theis, M. Enzymatically prepared RNAi
libraries. Nat. Methods 3, 696–700 (2006).
89. Bernards, R., Brummelkamp, T.R. & Beijersbergen, R.L. shRNA libraries and
their use in cancer genetics. Nat. Methods 3, 701–706 (2006).
90. Chang, K., Elledge, S.J. & Hannon, G.J. Lessons from nature: microRNA-based
shRNA libraries. Nat. Methods 3, 707–714 (2006).
91. Root, D.E., Hacohen, N., Hahn, W.C., Lander, E.S. & Sabatini, D.M. Genome-
scale loss-of-function screening with a lentiviral RNAi library. Nat. Methods
3, 715–719 (2006).
92. Pillai, R.S. MicroRNA function: multiple mechanisms for a tiny RNA? RNA 11,
93. Yekta, S., Shih, I.H. & Bartel, D.P. MicroRNA-directed cleavage of HOXB8
mRNA. Science 304, 594–596 (2004).
94. Rand, T.A., Petersen, S., Du, F. & Wang, X. Argonaute2 cleaves the anti-guide
strand of siRNA during RISC activation. Cell 123, 621–629 (2005).
95. Leuschner, P.J., Ameres, S.L., Kueng, S. & Martinez, J. Cleavage of the
siRNA passenger strand during RISC assembly in human cells. EMBO Rep. 7,
676 | VOL.3 NO.9 | SEPTEMBER 2006 | NATURE METHODS
© 2006 Nature Publishing Group http://www.nature.com/naturemethods