New Algorithm for Analysis of Off-target Effects in siRNA Screens

Karol Kozak

, Sandra Kaestner

, Thomas Wild

, Andreas Vonderheit

, Benjamin Misselwitz

Ulrike Kutay

and Gabor Csucs

LMSC, ETH Zurich, Schafmattstr. 18, 8093 Zurich, Switzerland

Institute of Biochemsitry, ETH Zurich, Schafmattstr. 18, 8093 Zurich, Switzerland

Institute of Microbiology, ETH Zurich, Schafmattstr. 18, 8093 Zurich, Switzerland

Keywords: RNAi, siRNA, Off-target, Transcriptome, Bioinformatics, Cell based Screening.

Abstract: The occurrence of RNAi side effects called “off-target effects” is still a challenging aspect in the

interpretation of data from large-scale RNA interference screens. To reduce off-target effects, improved

algorithms have been developed for small interfering RNA (siRNA) design, but also chemical modifications

of double stranded RNA molecules were introduced by the various commercial providers. To aid the

analysis of large-scale screens, we present a new algorithm and tool for the prediction of potential off-target

effects that can be applied to RNAi experimental data. Our approach provides different possibilities to

search for homologies between individual siRNAs of cellular mRNAs. We demonstrate our approach on a

ribosomal RNAi screening dataset.

1 INTRODUCTION

RNA interference (RNAi) has become a powerful

method for post-transcriptional silencing of specific

genes (Hannon, 2002). RNAi comprises different

small RNA molecules, which all make use of the

RNA-induced silencing complex (RISC) in order to

knock-down proteins. The current paper

concentrates on siRNA (small inhibitory RNA) (see

Figure 1). These RNA molecules are double

stranded RNA of a length of 21 bp (basepairs). The

two strands are called antisense (active, guide) and

sense (inactive, star), whereas the antisense strand

will bind to the corresponding mRNA. The first 2 –

8 bases of the antisense siRNA is called seed region

and at bases 8 – 10 is the cleavage site.

For knock-down/screening purposes different

companies offer sets of siRNAs targeting the whole

genome (or a subset of it) for various organisms.

Typically, they offer at least three different siRNAs,

for each target gene. These siRNAs can be used

either as single siRNA or can be mixed and used as a

pool of siRNAs. In this paper we concentrate on

single siRNAs and don’t deal with the specific

issues connected to pooling. The main reason for

offering several siRNAs per target is the varying

knock-down efficiency of the individual oligos and

the occurrence of off-target effects (that can have a

number of biological reasons). In our study, we

focus on sequence-dependent off-target effects that

can be attributed to the binding of the siRNA to

other mRNAs than their target mRNA (Fedorov et

al., 2005); (Jackson et al., 2006); (Fedorov et al.,

2006). This effect is caused by a high degree of

sequence complementarity/similarity. The

specificity of the siRNA sequence is thus a crucial

factor in an RNAi experiment (Semizarov et al.,

2005). Gene expression silencing through the RNAi

machinery works perfectly if the siRNA is totally

complementarity to its target mRNA. Single

nucleotide mismatches between the siRNA and the

target mRNA decrease the rate of mRNA

degradation (Haley and Zamore, 2006); (Elbashir et

al., 2001). The algorithms of the different companies

for generating the best siRNA sequence typically

take this into account and check and exclude siRNA

sequences that have total complementarity to other

than the target mRNA. However, also partial

complementarity between siRNA and mRNA seems

to result in a silencing effect (Jackson et al., 2006).

Based on this tolerance, siRNAs could have up to

hundreds of potential target sequences in the

genome. Currently, the degree of complementarity

between the two sequences needed for silencing is

not well defined.

Sequence-dependent off-target effects can be

253

Kozak K., Kaestner S., Wild T., Vonderheit A., Misselwitz B., Kutay U. and Csucs G..

New Algorithm for Analysis of Off-target Effects in siRNA Screens.

DOI: 10.5220/0004185802530260

In Proceedings of the International Conference on Bioinformatics Models, Methods and Algorithms (BIOINFORMATICS-2013), pages 253-260

ISBN: 978-989-8565-35-8

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

caused by a number of mechanisms and they are

summarized in Table 1.

First of all, it has been reported that off-target

effects occur with a high probability, if the siRNA

shows ~90% complementarity (17 nucleotides out of

19) to an off-target gene (Birmingham et al., 2006);

(Jackson et al., 2003). However, a 21-nucleotide

double-stranded RNA sharing only partial

complementarity with an mRNA is still competent to

cause gene silencing via translational repression

(Saxena et al., 2003); (Jackson et al., 2003). It seems

that already as few as 11 contiguous

complementarity nucleotides or a total of 15 are

sufficient to reduce the level of mRNA transcripts

(Jackson et al. 2003). The complementarity of the

siRNA seed region (the first 2–8 bases of the

antisense siRNA-strand) plays a major role in the

occurrences of off-target effects (Jackson et al.,

2006) (see Figure 1). Further analyses showed a high

tolerance for mismatches outside of the seed region,

whereas differences within this 5’ end of the siRNA

are barely tolerated (Amarzguioui et al., 2007);

(Lewis et al., 2005); (Doench et al., 2003).

The center region of the siRNA is important to

stabilize the siRNA-mRNA-duplex and to enhance

mRNA degradation (Saxena et al., 2003). Alemán

and colleagues analysed this central region, which

comprises the cleavage site of the mRNA (position

8-10 of the antisense strand, see Figure 1). They

deciphered that mismatches in this region of the

siRNA seem to be critical (Alemán et al., 2007) and

result in no cleavage. Additionally, they also tested

the aspect of a G:U wobble and discovered that the

G:U base pair is recognized like an authentic

Watson-Crick base pair in the anti-sense RNA-

mRNA duplex. This wobble base pairing expands

the range of potential targets for a specific siRNA.

Figure 1: Structure of an siRNA: 21 bp RNA duplex with

2 nucleotide 3’ overhang on each strand, the two strands

are called antisense or active or guide strand and sense or

inactive or star strand, respectively, the first 2 – 8 bases of

the antisense strand are called seed region and at bases 8 –

10 of the antisense siRNA strand is the cleavage site.

Anderson et al., calculated the frequencies at

which the seed hexamers appear in the 3’UTR

transcriptome and called this the seed

complementarity frequency (SCF) (Anderson et al.,

2008). They discovered that siRNAs with a low SCF

also have a low probability of generating an off-

target effect. Finally, Ui-Tei et al. found that the

seed-dependent off-target effect is highly correlated

with the thermodynamic stability in the duplex

formed between the seed region of the siRNA guide

strand and its target mRNA (Ui-Tei et al., 2008).

Off-target effects seem to occur if there exists a high

thermodynamic stability in the 5’ region caused for

example through a high G/C content (Lin et al.,

2005). This leads to the conclusion that it is not the

absolute number of mismatches but probably the

overall stability of the siRNA-mRNA duplex what

determines the success of a silencing event.

In order to predict off-target effects, a number of

methods or algorithms can be applied. For example,

the Basic Local Alignment Search Tool (BLAST)

(Altschul et al., 1990) is adopted to find nearly exact

homologies. Although BLAST is an excellent tool

for broad sequence alignments, it falls short in its

ability to accurately predict small local homologies.

Other bioinformatics tools (Alemán et al., 2007; Lin

et al. 2005), which don’t have this shortcoming, try

to predict interactions between siRNAs and mRNAs.

But unfortunately, these sequence-based prediction

tools frequently don’t consider specific off-target

parameters like: target site location, 3’ UTR

conserved regions, design specificity. Here, we

describe a novel method and a software

environment, supporting the analysis of potential

off-target effects for every siRNA of interest. This

environment enables researchers to determine

potential off-target effects in high throughput siRNA

experimental results.

2 METHODS

2.1 Concept for a Bioinformatics

Analysis of Off-target Effects

Available sequence analysis tools fail to reliably

predict off-target effects for siRNA sequences.

Building upon current understanding for the

occurrence of off-target effects, a new modular

analytic process is introduced here. This process can

be specifically adapted to a variety of options in

results interpretation to identify potential off-target

genes for every siRNA of interest. In the next

subsection, different scenarios for a meaningful

application of such an off-target analysis are

elaborated.

To verify the results of an RNAi screen that

relied on e.g. 4 different siRNAs per target the

screening results are evaluated with respect to how

many of the 4 siRNA per target gave rise to a

BIOINFORMATICS2013-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

254

phenotype. If all four siRNAs show an effect of

similar magnitude, one can reasonably be sure that

the downregulation of the intended target gene had

worked. However, most often, not all siRNAs for

one target yield a similar phenotype, and frequently,

the 4 siRNAs show a graded response in the

respective assay and sometimes only one or two

siRNAs show an effect, whereas the others don’t

show any effect. Very often 3 out 4 siRNAs are also

consider as a strong candidate. By performing the

off-target analysis, one can determine whether for

such siRNA there exists such a complementarity

with a top hit or not.

Table 1: Cause for sequence-dependent off-target effects.

1 Causes for off-target effects References

2 nearly exact complementarity

(Jackson et al., 2003;

Birmingham et al., 2006)

15 nucleotides (nt) in total or even 11

continuous nt match

(Jackson et al., 2003)

4 seed region complementarity

(Jackson et al., 2006;

Lin et al., 2005;

Amarzguioui et al., 2003)

miRNA function (seed region-3’UTR

conserved region complementarity)

(Doench et al. 2003)

multiple occurrences of the seed

region in an mRNA sequence

(Doench et al., 2003;

Lin et al., 2005)

complementary region at the

cleavage site, center of the siRNA

(Alemán et al., 2007;

Saxena, 2003)

8 tolerance of G:U wobble (Alemán et al., 2007)

9 seed complementation frequency (Amarzguioui et al., 2003)

10 high G/C content in the seed region (Lin et al., 2005)

2.2 The Analytic Process

Potential off-target effects are predicted based on

sequence complementarity regions between siRNAs

and mRNAs. For flexibility and extensibility

reasons, the process is composed as a set of steps,

which must be performed in sequence to get to an

effective analysis (see Figure 2).

After selecting the potential off-target siRNAs,

the next step is to find homologies between these

siRNAs and all mRNAs. This concept contains

many variants for such a complementarity search

using different algorithms to perform a sequence

alignment between siRNA and mRNA. A detailed

description of the different complementarity search

strategies is given in the next subsection.

The resulting list of a complementarity search

can be too long to find the important results just by

visual inspection. Therefore the next step is to filter

this list to reduce its size to meaningful results (see

section filter options).

2.2.1 Complementarity Search

In this analysis step, it can be determined if there

exists a complementary region between the selected

siRNA sequences and the mRNAs. Many different

sequence alignment algorithms are available to

perform such a complementarity search, but they are

not optimal for the purpose of this process step by

default. Therefore, three different strategies for the

use of these algorithms have been developed to find

nearly exact complementary regions as well as small

local complementarities (see also Figure 2).

Figure 2: General structure of the concept for analysing

screening results of off-target effects. Three variants for a

complementarity search to find potential off-target effects,

which correspond to type of off-targets in Table1: lines 3,

5, 8, 9.

2.2.2 BLAST Search

The Basic Local Alignment Search Tool (BLAST)

(Altschul et al., 1990) is one of the most popular

algorithms for complementarity search and can be

applied to find nearly identical gene regions for a

specific siRNA sequence. Unfortunately, the

BLAST algorithm is not applicable to find all

potential off-target candidates, because genes with

only partial complementarity will be missing. A

complementarity between the seed region (positions

2-7 on the antisense strand) of the siRNA and the

mRNA sequence might be sufficient to cause off-

target effects. But initially, BLAST looks for

continuous matches that are at least 7 nt long and

thus would overlook genes with only a seed region

complementarity (Table 1: lines 2 and 3). Despite

this disadvantage, BLAST is an effective tool to find

out immediately if obvious off-target genes exist

with a nearly identical nucleotide sequence to the

siRNA (Table 1: line 1).

In contrast to BLAST, other local alignment

NewAlgorithmforAnalysisofOff-targetEffectsinsiRNAScreens

255

algorithms can find small partial complementarities

between siRNA and mRNA sequences. Therefore,

the developed concept offers, besides the BLAST

search, two different alternatives of building a local

alignment without getting into the runtime problem.

2.2.3 Smith-Waterman Algorithm

The Smith-Waterman algorithm is an accurate

algorithm used to build local alignments between

two sequences (Smith and Waterman, 1981). Since

its use with all mRNAs from the database is not

practicable, a feasible alternative is to limit the

number of mRNAs to approximately 200. The set of

200 mRNAs are mRNAs that have highest

alignment score with the siRNA. On a Windows 7

system (two Intel Xeon Quad-core 2.00 GHz CPUs,

16 GB RAM), the analysis of a 20 genes (4

oligonucleotides) library constructs took about 1

hour. By reducing the number of sequences it is

possible to perform a local alignment for all the

siRNAs.

2.2.4 Seed-Motif-Search Combined with the

Smith-Waterman Algorithm

Because of the mentioned runtime problem when

performing a local alignment with the Smith-

Waterman algorithm, a third variant to search for

complementarity is introduced here. In this variant,

an initial step reduces the length of the mRNA

sequences to enable the use of a local alignment

algorithm. This reduction is made because the seed

region of the siRNA seems to play a significant role

in causing off-target effects. At the beginning, all

occurrences of the seed motif of every siRNA are

localized in the genes (see Figure 3). After detecting

this small region, a sequence of ~50 nt around this

seed motif is cut out in the mRNA. Thus, as a result

of this first step, a huge number of sequences of ~50

nt in length are obtained containing the seed region

of each siRNA. Due to the small length of the

sequences it is now possible to perform a local

alignment with the Smith-Waterman algorithm.

2.2.5 Filter Options

The result of the complementarity search is a list of

potential off-target candidates. This list can be very

large including many false positives. To reduce its

size and to get only the most probable off-target

candidates, a great number of filtering options are

provided. The following filter options present a

central part of the analysis concept (see also Figure

3):

• Threshold filter: Only alignments that are higher

or lower than a specific threshold alignment score

• Position of matches in the alignment: The

positions at which the alignment should contain a

match can be specified with this filter. It can be

used, for example, to show only alignments

matching at position 9-11 in the siRNA sequence,

because this central region seems to play an

important role in the occurrence of off-target effects

(Table1:line 6). Optionally also a G:U wobble can

be tolerated as a match in the alignment (Table 1:

line 7).

• Number of matches in the alignment: The total

number of matches that should at least occur in the

alignment is defined in this filter (Table 1:line 8).

• Length of a continuous match in the alignment:

With this filter, the length of a continuous match can

be determined, e.g. the occurrence of a stretch of at

least 11 bases in the alignment (Table 1: line 2).

• Location within the mRNA: The location of the

alignment within the mRNA is specified in this

filter. For off-target it is important that siRNA

should match in the 3’UTR of the mRNA (Table 1:

line 3).

• Location within the 3’ UTR conserved region:

The conserved regions within 3’ UTRs of human

mRNAs are collected from UTRdb (Grillo et al.

2010).

• Multiple complementary regions of a siRNA in

one mRNA: Only the results of a siRNA which has

multiple complementary regions in the same mRNA

are shown (Table 1: line 5).

• Strand selection: This filter extracts

complementarities between the gene and either the

sense or the antisense strand of the siRNA.

• Specific gene or a list of genes: The list of off-

target candidates can be filtered for one or more

genes of interest. Such genes could be for example

the known top hits of the screen.

Figure 3: General schema the workflow representing

Filtering possibilities for off-target candidates through

different options which correspond to type of off-targets in

Table 1 ines 1, 2, 4, 5, 6, 7.

BIOINFORMATICS2013-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

256

Based on this objective, a software toolkit to

analyse RNAi screening data for off-target effects

has been implemented. This toolkit was integrated

into the High Content Data Chain (HCDC)

environment, which is an extension of the Konstanz

Information Miner (KNIME) platform. The main

aim of the implementation was to develop nodes as

flexible and user-friendly as possible. Consequently,

the software application to analyse off-target effects

implemented in HCDC is presented as a set of

nodes.

2.3 Workflows - Execution of

Individual Analyses

The methods and tools described above in

combination with the existing nodes provided from

HCDC and KNIME represent the toolkit for creating

individual workflows. To demonstrate how an

analysis of off-target effects can be performed, an

exemplary workflow is shown in Fig. 4. Node

settings used for this workflow are described on

webpage (HCDC).

Figure 4: Filter combinations in a workflow to analyze

off-target effects.

The filtering steps are of great importance in the

analytic process, because they help to remove false

positives from the output table and to find the most

probable off-target genes. As mentioned before, the

filter nodes can be used and combined completely

flexibly. This is demonstrated in Figure 4 showing

three sample workflows for performing meaningful

filtering.

It has been reported that siRNAs can cause off-

target effects because of their miRNA-like

behaviour (Doench et al., 2003). This means that

there is a seed region complementarity with the

3’UTR of the mRNA. The effect could be amplified

if this complementary region occurs multiple times

in the 3’ end of the mRNA. To filter for these off-

target effects, the filter nodes shown in variant A of

Figure 4 can be connected. The Alignment Filter

gets the results which show a stretch of matches at

the seed region (positions 13-18). Afterwards, the

Location Filter reduces the table to those alignments

located in the 3’UTR of the mRNA, and finally the

Multiple Hit Filter shows only the results where a

siRNA has at least 2 complementary regions in an

mRNA. This

example demonstrates the advantage of

the workflow environment. The filtered output tables

are always available for each node enabling the user

to compare the results after each filtering step.

Variant B in Figure 4 shows that also combinations

of the same type of filtering node are allowed. The

first Alignment Filter, for example, only displays the

alignments, which contain a match at the positions

9-11 and 13-18. Additionally, the second Alignment

Filter node reduces this output table to those

alignments having at least 11 matches in total. As a

last example, a very common filter combination to

analyse off-target effects is shown in variant C of

Figure 4. Off-target effects triggered by a

complementary region with a top hit gene of the

screen are normally of great interest. The occurrence

of such off-target effects would explain unexpected

phenomena in screening results. Therefore, the Text

Filter node can be used with a list of top hit gene

symbols so that only siRNA results showing a

complementary region with a top hit are included in

the output table afterwards. However, not all of

these results are correct off-target candidates and

further filtering nodes have to be applied to get the

most probable ones.

3 RESULTS

For the purpose of validating the new toolkit and

showing its usability in actual research projects, it

has been applied to experimental data from High

Content Screenings technology. The first dataset

used for validation is a RNAi screen for components

involved in ribosome biogenesis. Ribosomes are

macromolecular complexes that synthesize proteins.

Ribosomes are composed of a small and a large

subunit, which both consist of ribosomal proteins

and RNA. In eukaryotes, the biogenesis of these

subunits is a complex multistep process including

the assembly of different components into the

subunits in the nucleolus, the export of these

precursors to the cytoplasm, and final cytoplasmic

maturation steps. In this project, the biogenesis of

the small ribosomal subunit (40S subunit) was

studied in human cells by performing a genome-

wide siRNA screen. The Rps2-YFP assay used

(Wild et al., 2010) enables the visual detection of

nuclear 40S maturation defects upon depletion of a

NewAlgorithmforAnalysisofOff-targetEffectsinsiRNAScreens

257

Table 2: First test set applied to the Seed-Motif-Search Node (default parameters), the Alignment filter (matches at position

5-18) and the Threshold Filter (-94). The first 4 columns describe: 1- gene name, 2- GeneID, 3- RefSeq, 4- gene

description. The last 3 columns provide the alignment score of siRNA and mRNA and the position of the alignment on the

mRNA.

protein by RNAi. As a numerical readout, a hit rate

was determined for each siRNA, measuring the

amount of cells displaying ribosome synthesis

defects. In total, 17632 genes and 5318 predicted

genes where each gene is target by four oligos.

In a first step, 13 siRNAs of the first dataset

were selected such that one siRNA has a hit rate >

0.9, while the other three siRNAs targeting this gene

have a hit rate < 0.03. Applying these criteria yields

many genes, which constitute potential off-target

hits. Less stringent filtering criteria may be applied

to get a longer list of potential off-target hits.

However, for the purpose of this software validation

analysis, it seems beneficial to visualize the off-

target analysis of a small set of genes. To find

complementary regions between the siRNAs

(targeting the selected genes) and all human

mRNAs, the Seed-Motif-Search is used. Since the

resulting list has a size of 477968 rows, a large

number of filter combinations is applied afterwards

to reduce its size and to detect the most probable off-

target candidates for an siRNA. Depending on the

combination of different filter nodes the possible

off-target genes vary considerably. An example of a

strict filtering is the use of the Alignment Filter

which allows only alignments containing matches at

position 5-18. The output of this node is shown in

Table 2.

All obtained potential off-target genes seem to be

good off-target candidates since they show a high

percentage of complementarity with the respective

siRNA.

For one gene (DTNBP1) a potential off-target is

RPS3, a ribosomal protein required for 40S

biogenesis and a top hit in the provided dataset.

Hence, an off-target effect of the siRNA

targeting DTNBP1 against RPS3 might explain the

high hit rate observed for this siRNA.

There is no fixed score that would give a clear

suggestion for potential off-target. Nevertheless,

after several observations and analyses we can

suggest that an alignment score above 73 with a

match length minimum of 14 nt will predict a

potential off-target with a reasonable fidelity.

3.1 Validation

To experimentally test the off-target prediction

results (Table 2), we analysed RPS3 levels upon

RNAi against DTNBP1. Using the same readout as

in the screen (nuclear accumulation of Rps2-YFP),

we tested all 4 siRNAs against DTNBP1 present in

the genome-wide screen. Additionally, we measured

RPS3 levels by Western blotting (Figure 5). As

observed in the genome-wide screen, si-DTNBP-2

treatment leads to 40S biogenesis defects

comparable to si-RPS3 treatment. In contrast, the

other siRNAs against DTNBP1 cause no 40S

biogenesis defect (comparable to untreated and si-

control sample). Importantly, RPS3 levels were

significantly reduced upon si-DTNBP1-2 treatment,

but unaffected by the other siRNAs against

DTNBP1. Therefore, the high hit rate of si-

DTNBP1-2 is likely caused by RPS3 depletion and

hence, as predicted, an off-target effect.

Figure 5: Experimental verification of predicted off-target

effect.

A HeLa Rps2-YFP cells were treated with

indicated siRNAs according to Wild et al. Images

were taken with epi-fluorescence microscopy.

BIOINFORMATICS2013-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

258

B Protein levels of RPS3 and ENP1 (as a loading

control) derived from experiment shown in A were

detected by Western blotting. Note that si-DTNBP1-

2 reduces RPS3 levels.

4 CONCLUSIONS

The major objective of this paper was to describe a

new algorithm and software toolkit to analyse off-

target effects in RNAi screening data.

The test and validation phase has proven that the

software already provides a powerful and flexible

toolkit for analysing off-target effects. Testing and

validating the toolkit with an actual high content

dataset revealed that the workflow environment is

suitable for off-target analysis. The analyses of the

given dataset show that the environment allows for a

dynamic workflow adaption based on intermediate

results, e.g. by supplemental Text Filter integration.

In addition to the flexible workflow creation facility,

the individual configuration options of a single node

are also advantageous. All in all, the software

environment with its flexibility turns out to be very

suitable to analyse off-target effects in RNAi

screening data. An important aspect is the reliability

of the results obtained in the analysis process. In this

case the results seem to be reasonable and correct.

We would like emphasize that our predictions

neither include the effects of siRNA concentration

nor do they attempt to account for the siRNA pool

constructs. It is clear that both these effects are of

critical practical consequence and that a

computational model supporting them is desirable.

At the moment, however, there is insufficient

published data on the efficacies of pools to be able

to construct a high-confidence model of pool effects.

REFERENCES

Alemán, L. M., Doench, J. G. and Sharp, P. A., 2007.

Comparison of sirna-induced off-target rna and protein

effects. RNA, 13(3):385–95.

Altschul, S. F., Gish, W., Miller, W., Myers, E. W and

Lipman, D. J., 1990. Basic local alignment search tool.

J Mol Biol, 215(3):403–10.

Amarzguioui, M., Holen,T., Babaie, E. and Prydz, H.

2003. Tolerance for mutations and chemical

modifications in a sirna. Nucleic Acids Res,

31(2):589–95.

Anderson, E. M., Birmingham, A., Baskerville, S.,

Reynolds, A., Maksimova, E., Leake, D., Fedorov, Y.,

Karpilow, J. and Khvorova, A., 2008. Experimental

validation of the importance of seed complement

frequency to sirna specificity. RNA, 14(5):853–61.

Birmingham, A., Anderson, E. M., Reynolds, A., Ilsley-

Tyree, D., Leake, D., Fedorov, Y., Baskerville, S.,

Maksimova, E., Robinson, K., Karpilow, J., Marshall,

W.S and Khvorova, A., 2006. 3’ utr seed matches, but

not overall identity, are associated with rnai off-

targets. Nat Methods, 3(3):199–204.

Doench, J. G., Petersen, C. P. and Sharp, P. A., 2003.

Sirnas can function as mirnas. Genes Dev, 17(4):438–

42.

Elbashir, S. M, Martinez, J., Patkaniowska, A., Lendeckel,

W. and Tuschl, T., 2001. Functional anatomy of sirnas

for mediating efficient rnai in drosophila melanogaster

embryo lysate. EMBO J, 20(23):6877–88.

Fedorov, Y., King, A., Anderson, E. M., Karpilow, J.,

Ilsley, D., Marshall, W. S. and Khvorova, A., 2005.

Different delivery methods-different expression

profiles. Nat Methods, 2(4):241.

Fedorov, Y., Anderson, E. M, Birmingham, A., Reynolds,

A., Karpilow, J., Robinson, K., Leake, D., Marshall,

W.S. and Khvorova, A., 2006. Off-target effects by

sirna can induce toxic phenotype. RNA, 12(7):1188–

96.

Grillo, G., Turi, A., Licciulli, F., Mignone, F., Liuni, S.,

Banfi, S., Gennarino, V. A., Horner, D. S., Pavesi, G.,

Picardi, E., Pesole, G., 2010. UTRdb and UTRsite

(RELEASE 2010): a collection of sequences and

regulatory motifs of the untranslated regions of

eukaryotic mRNAs. Nucleic Acids Res., 38, 75-80.

Hannon, H. J., 2002. Rna interference. Nature,

418(6894):244–51, July.

Haley, B. and Zamore, P. D., 2004. Kinetic analysis of the

rnai enzyme complex. Nat Struct Mol Biol, 11(7):599–

606.

HCDC web page, bioinformatics module:

[http://hcdc.ethz.ch/index.php?view=article&id=25]

Jackson, A. L, Burchard, J., Leake,D., Reynolds, A.,

Schelter, J., Guo, J., Johnson, J. M., Lim, L.,

Karpilow, J., Nichols, K., Marshall, W. S., Khvorova,

A. and Linsley, P. S., 2006. Position-specific chemical

modification of sirnas reduces "off-target" transcript

silencing. RNA, 12(7):1197–205.

Jackson, A. L., Burchard, J., Schelter, J., Chau, B. N.,

Cleary, M., Lim, L. and Linsley, P. S., 2006.

Widespread sirna "off-target" transcript silencing

mediated by seed region sequence complementarity.

RNA, 12(7):1179–87.

Jackson, A. L., Bartz, S. R., Schelter, J., Kobayashi, S. V.,

Burchard, J., Mao, M., Li, B., Cavet, G. and Linsley,

P. S., 2003. Expression profiling reveals off-target

gene regulation by rnai. Nat Biotechnol, 21(6):635–7.

Lewis, B. P., Burge, C. B. and Bartel, D. P., 2005.

Conserved seed pairing, often flanked by adenosines,

indicates that thousands of human genes are microrna

targets. Cell, 120(1):15–20.

Lin, X., Ruan, X., Anderson, M. G., McDowell, J. A.,

Kroeger, P. E., Fesik, S. W. and Shen, Y., 2005. Sirna-

mediated off-target gene silencing triggered by a 7 nt

complementation. Nucleic Acids Res, 33(14):4527–35.

NewAlgorithmforAnalysisofOff-targetEffectsinsiRNAScreens

259

Saxena, S., Jónsson, Z.O. and Dutta, A. 2003. Small rnas

with imperfect match to endogenous mrna repress

translation. implications for off-target activity of small

inhibitory rna in mammalian cells. J Biol Chem,

278(45):44312–9.

Semizarov, D., Frost, L., Sarthy, A., Kroeger, P.A,

Halbert, D. N. and Fesik, S. W., 2005. Specificity of

short interfering rna determined through gene

expression signatures. Proc Natl Acad Sci U S A,

100(11):6347–52.

Smith, T. F. and Waterman, M. S., 1981. Identification of

common molecular subsequences. J Mol Biol,

147(1):195–7.

Ui-Tei, K., Naito, Y., Nishi, K., Juni, A. and Saigo, K.,

2008. Thermodynamic stability and watson-crick base

pairing in the seed duplex are major determinants of

the efficiency of the sirna-based off-target effect.

Nucleic Acids Res, 36(22):7100–9.

Wild, T., Horvath, P., Wyler, E., Widmann, B.,

Badertscher, L., Zemp, I., Kozak, K., Csusc, G., Lund,

E., Kutay, U., (2010) A protein inventory of human

ribosome biogenesis reveals an essential function of

Exportin 5 in 60S subunit export. PLoS Biol,

8:e1000522.

BIOINFORMATICS2013-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

260