A Novel Pipeline for Identiﬁcation and Prioritization of Gene Fusions in

Patient-derived Xenografts of Metastatic Colorectal Cancer

Paciello Giulia

, Andrea Acquaviva

, Consalvo Petti

, Claudio Isella

, Enzo Medico

2,3

and Elisa Ficarra

Department of Control and Computer Engineering, Politecnico di Torino, Corso Duca degli Abruzzi, Turin, Italy

Laboratory of Oncogenomics, Institute for Cancer Research at Candiolo (IRCC),

Strada Provinciale 142, Candiolo, Turin, Italy

Department of Oncology, Universita’ di Torino, Candiolo, Turin, Italy

Keywords:

RNA-sequencing, Metastatic Colorectal Cancer, Patient-derived Mouse Xenografts, Artifacts.

Abstract:

Metastatic spread to the liver is a frequent complication of colorectal cancer (CRC), occurring in almost

half of the cases, for which personalized treatment strategies are highly desirable. To this aim, it has been

proven that patient-derived mouse xenografts (PDX) of liver-metastatic CRC can be used to discover new

therapeutic targets and determinants of drug resistance. To identify gene fusions in RNA-Seq data obtained

from such PDX samples, we propose a novel pipeline that tackles the following issues: (i) discriminating

human from murine RNA, to ﬁlter out transcripts contributed by the mouse stroma that supports the PDX;

(ii) increasing sensitivity in case of suboptimal RNA-Seq coverage; (iii) prioritizing the detected chimeric

transcripts by molecular features of the fusion and by functional relevance of the involved genes; (iv) providing

appropriate sequence information for subsequent validation of the identiﬁed fusions. The pipeline, built on top

of Chimerascan(R.Iyer, 2011) and deFuse(McPherson, 2011) aligner tools, was successfully applied to RNA-

Seq data from 11 PDX samples. Among the 299 fusion genes identiﬁed by the aforementioned softwares,

ﬁve were selected since passed all the ﬁltering stages implemented into the proposed pipeline resulting as

biologically relevant fusions. Three of them were experimentally conﬁrmed.

1 INTRODUCTION

It is currently known that cancer derives from per-

manent alterations of the cellular DNA, leading to

aberrant growth, invasion of adjacent tissues and

metastatic diffusion at distant sites(Hanahan, 2000).

Among the various DNA alterations, chromosomal

rearrangements leading to gene fusions play a cen-

tral role in the initial steps of many pathologies

such as leukaemias, sarcomas and common epithelial

neoplasms, like breast, colorectal and prostate can-

cer(Aman, 1999). The impact of gene fusions on cel-

lular behavior is due to functional alteration of one

or both the genes involved in the chromosomal re-

arrangements that give rise to chimeric transcripts.

Typical consequences of gene fusions, at the RNA

and protein level, are strong variation of expression,

removal of regulatory domains, forced oligomeriza-

tion, change of the subcellular location or acquisi-

tion of novel binding domains(Edwards, 2010). Gene

fusions can therefore have important prognostic and

therapeutic implications in the management of malig-

nancies, as shown in recent studies(Mitelman, 2007).

Data produced by Next Generation Sequenc-

ing (NGS) technologies are nowadays considered

very useful in order to detect genetic abnormali-

ties(Ansorge, 2009)(Mardis, 2008)(Metzker, 2010).

In particular, concerning chimeric transcripts iden-

tiﬁcation, the analysis of the so called paired-end

RNA-Seq reads can be considered a powerful strat-

egy as shown by Maher and colleagues(Maher, 2009)

and conﬁrmed by the subsequent development of nu-

merous tools to perform such activity(McPherson,

2011)(R.Iyer, 2011)(Abate, 2012). Paired-end reads,

differently from single-end reads, are obtained by se-

quencing nucleic acid fragments at both the 5’ and

3’ ends. When the two sequenced portions of the

fragment, called mates, align on different genes, it is

likely that the fragment is originated by a chimeric

transcript. We applied this strategy to identify rele-

vant gene fusions in colorectal cancer (CRC), one of

the most frequent cancers worldwide(Walther, 2000).

142

Giulia P., Acquaviva A., Petti C., Isella C., Medico E. and Ficarra E..

A Novel Pipeline for Identiﬁcation and Prioritization of Gene Fusions in Patient-derived Xenografts of Metastatic Colorectal Cancer.

DOI: 10.5220/0004799401420148

In Proceedings of the International Conference on Bioinformatics Models, Methods and Algorithms (BIOINFORMATICS-2014), pages 142-148

ISBN: 978-989-758-012-3

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

CRC is frequently complicated by liver metastasis,

and features a remarkable heterogeneity in terms of

molecular pathogenesis, natural history and response

to treatment(Siena, 2009)(Cunningham, 2010). A

recent advance in the characterization of such het-

erogeneity has been brought forward by the prop-

agation of human neoplastic tissue in immunodeﬁ-

cient mice, the so-called patient-derived xenograft

(PDX) approach. As proven by Bertotti and col-

leagues(Bertotti, 2011), PDXs of human metastatic

colorectal cancer can be reliably exploited to discover

novel determinants of therapeutic response and new

oncoprotein targets. PDXs are indeed able to con-

serve the inter-individual diversity and the genetic

heterogeneity typical of the tumors of origin and at

the same time to reproduce the disease responses in

humans.

In the present work we searched for chimeric tran-

scripts in eleven PDXs of metastatic CRC, by ana-

lyzing Illumina RNA-Seq data consisting of 100-base

pair (bp) long paired-end reads. RNA was extracted

from PDXs at the second passage of propagation in

mice. By this stage human stromal cells, not capa-

ble of growing in the murine context, are replaced

by mouse stromal cells. As a consequence, the ex-

tracted RNA is a mixture of human RNA, originated

from neoplastic cells, and murine RNA from stromal

cells. This required a dedicated step in the RNA-Seq

analysis pipeline to distinguish reads originated from

RNAs of the two species. The experimental setup is

reported in Figure 1.

Figure 1: Experimental Setup. The activities highlighted in

blue boxes are respectively that implemented in the pipeline

objective of this work and the validation of the obtained re-

sults.

In particular rhe proposed research is inspired by

the observation that fusion detection tools usually

produce in the preliminary steps of their workﬂow a

conspicuous number of putative gene fusions not fea-

sible for in lab validation. To shrink down these can-

didates lists, gene fusions detection tools implement

inside their algorithms different kind of ﬁlters gener-

ally efﬁcient in discarding false positive fusion tran-

scrpits but at the same time determining poor sensi-

tivity level of the detection. Therefore, as reported By

Abate and colleagues (Abate, 2012) and Carrara and

colleagues(Carrara, 2013) because the heterogeneity

of algorithms and ﬁlters applied, the identiﬁed fusions

usually poorly overlap with a high rate of false posi-

tives, but also of false negatives, leading to the need

of considering the union set of fusion genes detec-

tion tool outputs for in lab validation. The number

of fusions to be tested becomes in this way not feasi-

ble even considered that the biologically relevant fu-

sions in a sample are usually very few if any(Ozsolak,

2011).

With the aim of reducing the union set of candi-

dates detected by two fusion genes discovery tools (i.e

Chimerascan(R.Iyer, 2011) and defuse(McPherson,

2011)) in the aforementioned PDXs samples, we pro-

posed a novel pipeline characterized by the reimple-

mentation of some modules proper of fusion genes

detection tools and ad hoc scripts developed to per-

form different ﬁltering stages.

At the days no tools or algorithms implementing

the analysis performed by our pipeline can be iden-

tiﬁed since all sequencing studies in PDXs such as

those of Rossello and colleagues, Conway and col-

leagues, Valder and colleagues(Rossello, 2013)(Con-

way, 2012)(Valdes, 2013) circumscribed at the dis-

crimination between murine and human reads don’t

relying on fusion genes detection beyond their priori-

tizazion.

The whole workﬂow is implemented in order to:

(i) take account for the murine stroma; (ii) consider

the contingent PCR artifacts; (iii) evaluate the role of

different kind of reads in the chimeric transcript reli-

ability to maximize sensitivity also at low sequencing

coverage; (iv) integrate biological and functional in-

formation about the gene fusions. Among the priori-

tized fusions, three have been at the moment experi-

mentally conﬁrmed in lab with PCR.

2 METHODS

The proposed pipeline is characterized by different

activities and ﬁltering stages as shown in Figure 2.

In the following all the steps will be detailed.

First Filtering Stage: Gene Fusions Annotation

and Selection.

The list of gene fusions detected using Chimeras-

can(R.Iyer, 2011) and defuse(McPherson, 2011) tools

with default run parameters constitute the input of

the ﬁrst step of the proposed workﬂow on the eleven

RNA-Seq samples under examination. The choice of

Chimerascan(R.Iyer, 2011) and deFuse(McPherson,

ANovelPipelineforIdentificationandPrioritizationofGeneFusionsin

Patient-derivedXenograftsofMetastaticColorectalCancer

143

Figure 2: Pipeline workﬂow. Activities are represented in

yellow boxes whereas input and output in blue ones.

2011) tools is dictated by a recent research performed

by Carrara and colleagues(Carrara, 2013). Chimeras-

can(R.Iyer, 2011) and deFuse(McPherson, 2011)

tools were indeed proven to achieve good sensitivity

levels on real dataset even if generally provide a

remarkable number of false positive chimeras. This

negative feature has been however properly managed

thanks to the following different ﬁltering stages. Fur-

thermore for what is concerning Chimerascan(R.Iyer,

2011) and defuse(McPherson, 2011) run parame-

ters they can be conveniently triggered according

to the speciﬁc requirements even if we evaluated

that from a computational point of view it is more

convenient to impose restrictive mapping policies

in the following phases of the proposed pipeline.

Chimeric transcripts output ﬁles, containing all the

fusions detected in the samples, are automatically

elaborated taking advantage of an annotation tool

(http://sourceforge.net/projects/pegasus-fus/) able

to perform on the detected fusions, protein domain

and functional analysis. The achieved information

provide a more detailed overview on each chimeric

transcript supplying knowledge about the presence

of kinases, the reading frame, the domain conserved

or loss, and the breakpoint regions. Of essential

importance in this phase are also the data provided

by the chimeric transcripts detection tools for what is

concerning the reads used to identify a gene fusion.

Chimerascan(R.Iyer, 2011) and deFuse(McPherson,

2011) indeed, in order to score the fusions detected

in the samples, provide the number of reads used

to deﬁne the fusion: They distinguish between

encompassing reads, in other word paired-end reads

with the two mates mapped respectively on the two

different partner genes of the chimeric transcript and

split mates, representing those mates that harbor the

gene fusion breakpoint. Starting from the information

collected using the aforementioned annotation tool,

the gene fusions detection tools outputs and thanks

to biological considerations about the function of the

genes involved in the fusion a ﬁrst step of ﬁltering

is performed. Only those fusions characterized by a

certain threshold number of split reads that will be

discussed in Results Section and satisfying also at

least another one criteria of those previously listed

will be considered for further evaluations.

Second Filtering Stage: Breakpoint Sequences

Analysis and Gene Fusions Selection.

Starting from the prioritized list of fusion gene

candidates a new stage of ﬁltering is applied. The

fusion sequence of each chimeric transcript, provided

by defuse(McPherson, 2011) and deduced from

the split reads for what is concerning Chimeras-

can(R.Iyer, 2011) is here analysed. Main objective

of this ﬁltering stage consists essentially in the

retrieving of those gene fusions sequences that could

account for the translation of the chimeric transcript

into a functional protein. The presence of a Kozac

sequence in the 5’ partner gene generally account

for an ATG triplet (starting site of the translation

process) downstream that allows the beginning of

the translation process. In this phase also chimeric

transcript sequences characterized by an in frame

conﬁguration have been however selected for further

analyses since we considered the case in which the

ATG triplet is located upstream with respect to the

starting point of the sequencing.

Third Filtering Stage: Paired-end Reads Mapping

and Gene Fusions Selection.

For the gene fusions selected in the previous step

the mapping of the paired-end reads on the fu-

sion sequence is performed taking advantage of

Bowtie(Langmead, 2009) tool which parameters

were set in order to report for a data read only

the best alignment identiﬁed. Outputs were later

converted using the Samtools(Li, 2009) and the

BedTools(Aaron, 2009) utilities to obtain suitable

format ﬁles for the following phases of the pipeline.

Gene fusions not supported by a threshold number of

paired-end reads will be not considered in the next

steps of the ﬂow.

Fourth Filtering Stage: PCR and Mouse Mates

Filtering.

All the mates mapped onto a speciﬁc fusion sequence

have been analyzed in this phase with the main pur-

poses of removing those deriving from PCR artifacts

BIOINFORMATICS2014-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

144

and those mapping also on the murine DNA. For what

is concerning PCR artifacts, as depicts in Figure 3, if

more than a mate is mapped in the same location, only

one mate of the group is considered for further analy-

ses (green mates in Figure 3 a).

Figure 3: PCR artifacts removal ﬁlter. In Subﬁgure a is

reported an example of mates deriving from PCR artifacts,

characterized by the same start and end mapping positions

on the fusion sequence. Only one read for group (the green

ones) is maintained for subsequent analyses. In Subﬁgure

b are reported instead some mates mapped in unambiguous

ladder-like pattern on the fusion sequence. All the mates are

retained in this case for further examination.

As shown in Figure 4 a only the mates survived to

the PCR artifacts removal ﬁlter have been mapped

using Bowtie(Langmead, 2009) in single-end mode

on the mouse genome in order to identify those

mates mapped also on this reference. These mates,

that probably derives from murine stromal cells

sequencing (Figure 4 b), are removed from the list

of mates supporting the speciﬁc gene fusion (green

mates in Figure 4).

Figure 4: Mouse artifacts removal ﬁlter. In Subﬁgure 4 a are

shown the input mates for the murine ﬁlter (all the mates not

deriving from PCR artifacts) that have been mapped onto

the fusion sequence. Mates belonging to such a group that

have been mapped also on the murine genome, as shown in

Figure 4 b (green mates), are not considered in the following

phases of the pipeline.

Fifth Filtering Stage: Paired-end Reads Recon-

struction and Spanning Reads Identiﬁcation.

After the last described stage of ﬁltering each gene

fusion results supported by a list of single mates in-

stead of the initial paired-end reads obtained from

bowtie run on the fusion sequence. As proven by

Maher et colleagues(Maher, 2009) paired-end reads

are however more meaningful for gene fusions de-

tection so, starting from the list of single mates sup-

porting a putative chimeric transcript the paired-end

read, if present, is reconstructed (green reads in Fig-

ure 5). Mates not belonging to a paired-end reads are

discarded. Among these reads of great importance

are the spanning reads, those reads characterized by

the mapping of the mates on the two different partner

genes of the chimeric transcript. Only those paired-

end reads (spanning reads) having the two mates par-

tially or totally mapped onto the two different partner

genes were retrieved for further examination (yellow

reads in Figure 5). Gene fusions not supported by

a certain number of spanning reads that will be dis-

cussed in Results Section were discarded.

Figure 5: Paired-end and spanning reads representation. In

Figure are reported respectively in green and yellow colours

the paired-end and the spanning reads.

Sixth Filtering Stage: Split Reads Retrieving and

Interesting Fusions Identiﬁcation.

In this phase, in order to guarantee once again the reli-

ability of both the partner genes involved in the fusion

and the breakpoint coordinates on the same partner

genes, the search for split reads is performed on the

remaining putative chimeric transcripts: Split reads

are indeed capable to account for a base-pair reso-

lution of the gene fusion sequence in the breakpoint

region. This is the reason for which only those span-

ning couples, represented with yellow bars in Figure

6, having one or both the mates mapped in the break-

point region provided by Chimerascan(R.Iyer, 2011)

and deFuse(McPherson, 2011) tools are considered.

The mates belonging to such a couple are called split

mates and are represented in a dashed box in Figure 6.

PCR validation will be performed for those gene fu-

sions characterized at least by a threshold number of

split mates discussed once a time in Results Section.

Figure 6: Split mates representation. In Figure are reported

some spanning paired-end reads. The mates belonging to

such couples that are mapped in the breakpoint region are

called split mates and are represented in a dashed box.

3 RESULTS

The run of deFuse(McPherson, 2011) on the eleven

metastatic CRC samples detected 132 gene fusions

whereas that of Chimerascan(R.Iyer, 2011) 167. The

two different outputs didn’t show overlapping fu-

sions as was initially supposed considering the re-

ANovelPipelineforIdentificationandPrioritizationofGeneFusionsin

Patient-derivedXenograftsofMetastaticColorectalCancer

145

cent researches before mentioned(Abate, 2012)(Car-

rara, 2013).These results, not so conspicuous if con-

sidered the number of samples under examination and

the mean amount of gene fusions usually identiﬁed

in RNA-Seq samples, are essentially due to the poor

coverage of the samples. It is worth noting how-

ever that generally higher the coverage of the sam-

ples higher will be the number of detected gene fu-

sions with remarkably computational costs spent for

the detection. It is at the same time worth noting that

not all the 299 fusions can be tested in lab using PCR

for known economic and temporal restraints. The pro-

posed pipeline was therefore applied in order to pri-

oritize the identiﬁed 299 chimeric transcripts. In the

following we refer to the different chimeric transcripts

using couples of capital letters (actual gene names

cannot be disclosed as the biological results of this

research are currently under review).

For what is concerning the ﬁrst ﬁltering stage a

threshold of one split read was imposed: Thirteen

gene fusions were here selected because supported

by at least one split read and present at least one

of the other features described in Subsection First

ﬁltering stage: Gene fusions annotation and selec-

tion. The two thresholds were selected with the in-

tention to be as preservative as possible in evaluat-

ing the different candidates. These parameter val-

ues can be however tuned in order to satisfy spe-

ciﬁc needs. Furthemore even if among the initial 299

fusion genes none have been previously detected in

cancer samples, all the genes involved in the thire-

teen fusions are characterized by mutational states re-

lated to cancer development and progression (infor-

mation deriving from literature sources and COSMIC

database(Simon, 2010)). Three out of the thirteen

chimeras that passed the ﬁrst ﬁltering stage are more-

over characterized by a partner gene found to be fused

with other genes in cancer diseases.

For each of these chimeric transcripts, the fusion

sequence has been then retrieved and analyzed in or-

der to understand the biological mechanism at the

basis of the recombination. The criteria of the sec-

ond ﬁltering stage reduced the previous list to only

eight gene fusions characterized by the presence of a

Kozac sequence (or an ATG triplet) at 5’-end or by

an in frame conﬁguration. Also frame shifted con-

ﬁgurations could however be interesting in case of

tumor suppressors 3’ partner genes: In the proposed

pipeline this scenario has not been considered because

no oncogenic suppressor genes were detected among

the identiﬁed chimeric transcript partner genes.

In Table 1 are reported the results relative to the

third and fourth ﬁltering stages. In particular, a

threshold of at least one paired-end read was ﬁxed

in order to select a fusion for the next phases of the

pipeline being as much preservative as possible: Only

one gene fusion (i.e gene fusion G-H) was deleted

because not supported by paired-end reads as shown

in column 3 of Table 1. The differences among

the breakpoint sequence lengths (Column 2 of Ta-

ble 1) can be attributed to the fact that they depend

on the number of reads used to deﬁne the fusion

sequence. So higher the number of reads mapped

by the chimeric transcript discovery tool on the sup-

posed breakpoint sequence, higher will be the pro-

vided length of the same sequence and the probability

of ﬁnding with the propose pipeline paired-end reads

aligning on the same. The absence of mates removed

by the mouse remove ﬁlter, shown in Column 5 of

Table 1, conﬁrm the fact that effectively the samples

were composed exclusively of human tumor cells. On

the other and, instead, the PCR in the most of cases

caused a remarkably number of artifacts, as it is pos-

sible to note from Column 4 of Table 1.

After PCR and mouse mates removal, for each

of the remaining seven gene fusions the supporting

paired-end reads, if present, are reconstructed. This

activity is followed by the identiﬁcation of the so

called paired-end spanning reads if existing. The re-

sults are shown in columns 2 and 3 of Table 2.

Two of the previous seven gene fusions have been

removed because they are not supported by spanning

reads (i.e. gene fusions I-L and M-N). A threshold

of one spanning read was indeed imposed in order to

consider a gene fusion. The value selected derives, as

already largely discussed, from the desire to be very

preservative since the previous ﬁltering stages con-

cerning functional and biological properties of fusion

genes have been already capable as shown to remove a

conspicuous number of not functional chimeric tran-

scrips. It is worth noting however that it is possible

to set this parameter according to the speciﬁc require-

ments.

The fourth column of Table 2 reports instead the

number of split mates supporting the remaining ﬁve

chimeric transcripts. In the last ﬁltering stage a

threshold of at least one split mate was imposed in

order to consider a gene fusion for in lab validation.

Even for the split reads the value parameter was se-

lected, as already said in relation to spanning reads

threshold, in order to be as conservative as possi-

ble. Of the initial 299 gene fusions at the end of

the pipeline ﬁve were considered priority (i.e gene fu-

sions A-B, C-D, E-F, O-P and Q-R). Three out of ﬁve

have been actually validated in lab using PCR result-

ing as true gene fusions. In table 3 is reported a sum-

mary of the number of fusion genes obtained after the

application of the different ﬁltering stages.

BIOINFORMATICS2014-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

146

Table 1: Third and Fourth Filtering Stages results.

Gene Fusion Gene Fusion sequence length Nr of mapped paired-end Reads PCR Mates Mouse Mates

A-B 3668 767 529 0

C-D 127 66 74 0

E-F 168 1 0 0

G-H 139 0 0 0

I-L 599 58 28 0

M-N 311 10 4 0

O-P 1886 32 2 0

Q-R 531 86 58 0

Table 2: Fifth and Sixth Filtering Stages results.

Gene Fusion Nr of paired-end Reads Nr of Spanning paired-end reads Number of Split Mates

A-B 344 2 2

C-D 16 11 12

E-F 1 1 2

I-L 18 0 0

M-N 6 0 0

O-P 30 2 3

Q-R 43 5 8

Table 3: Number of chimeric transcripts obtained as output of the different FSs (Filtering Stages).

#Initial fusions #Fusions I FS #Fusions II FS #Fusions III FS #Fusions IV FS #Fusions V FS

299 13 8 7 5 5

4 CONCLUSIONS AND FUTURE

WORK

Starting from the 299 gene fusions detected by

Chimerascan(R.Iyer, 2011) and deFuse(McPherson,

2011) on the eleven metastatic CRC xenopatients

RNA-Seq samples, the proposed pipeline was able to

progressively reduce the list of the identiﬁed chimeric

transcripts. Five gene fusions in particular passed all

the ﬁltering stages and were considered to be rele-

vant. The in lab validation of three of these gene fu-

sions conﬁrmed the presence of a chimeric transcript

product proving that the developed pipeline is able to

identify true chimeric transcripts potentially associ-

ated to cancer onset and progression. Furthermore all

the activities performed within the pipeline have been

implemented considering the general features at the

basis of a real, productive and biologically relevant

gene fusion making this program capable to priori-

tize gene fusions from PDXs affected also by differ-

ent diseases. Surely the proposed methodology can be

transferred to other RNA-Seq dataset deriving from

PDXs even with different pathologies. All the activ-

ities performed within the pipeline have been indeed

implemented considering the general features at the

basis of a real, productive and biologically relevant

gene fusion making this program capable to prioritize

gene fusions from PDXs affected as said by different

diseases. We would like to underline that our aim is

just to apply as soon as possible the proposed method-

ology o other dataset in order to improve the ﬁltering

stages with new features and at the same time to make

very user friendly the entire tool.

Future works will aim at: (i) validate all the detected

fusions using also Real-Time PCR (RT-PCR), (ii) in-

tegrate in the proposed pipeline results from other

chimeric transcript detection tools, (iii) change the

setting of the implemented ﬁlters in order to evalu-

ate the ﬁltering stages performances, (iv) investigate

the functional role of the identiﬁed fusion transcripts,

(v) evaluate the occurrence of the detected fusions in

public RNA-seq dataset (TCGA).

REFERENCES

Aaron, R. (2009). Bedtools: a ﬂexible suite of utilities for

comparing genomic features. Bioinformatics.

Abate, F. (2012). Bellerophontes: an rna-seq data analysis

framework for chimeric transcripts discovery based on

accurate fusion model. Bioinformatics.

Aman, P. (1999). Fusion genes in solid tumors. Semin

Cancer Biology.

ANovelPipelineforIdentificationandPrioritizationofGeneFusionsin

Patient-derivedXenograftsofMetastaticColorectalCancer

147

Ansorge, W. (2009). Next-generation dna sequencing tech-

niques. New Biotechnology.

Bertotti, A. (2011). A molecularly annotated platform of

patient-derived xenografts (’xenopatients’) identiﬁes

her2 as an effective therapeutic target in cetuximab-

resistant colorectal cancer. Cancer Discovery.

Carrara, M. (2013). State-of-the-art fusion-ﬁnder algo-

rithms sensitivity and speciﬁcity. BioMed Research

International.

Conway, T. (2012). Xenome–a tool for classifying reads

from xenograft samples. Bioinformatics.

Cunningham, D. (2010). Monoclonal antibodies in the

treatment of metastatic colorectal cancer: a review.

Colorectal Cancer.

Edwards, P. (2010). Fusion genes and chromosome translo-

cations in the common epithelial cancers. The Journ.

of Pathology.

Hanahan, D. (2000). The hallmarks of cancer. Cell.

Langmead, B. (2009). Ultrafast and memory-efﬁcient align-

ment of short dna sequences to the human genome.

Genome Biology.

Li, H. (2009). The sequence alignment/map (sam) format

and samtools. Bioinformatics.

Maher, C. (2009). Chimeric transcript discovery by paired-

end transcriptome sequencing. PNAS.

Mardis, E. (2008). The impact of next-generation sequenc-

ing technology on genetics. Trends in genetics.

McPherson, A. (2011). defuse: An algorithm for gene fu-

sion discovery in tumor rna-seq data. PLOS Compu-

tational Biology.

Metzker, M. (2010). Sequencing technologies - the next

generation. Nature Reviews Genetics.

Mitelman, F. (2007). The impact of translocations and gene

fusions on cancer causation. Nature Reviews.Cancer.

Ozsolak, F. (2011). Rna sequencing: advances, challenges

and opportunities. Nat Rev Genet.

R.Iyer (2011). Chimerascan: a tool for identifying chimeric

transcription in sequencing data. Bioinformatics.

Rossello, F. (2013). Next-generation sequence analysis of

cancer xenograft models. PLoS ONE.

Siena, S. (2009). Biomarkers predicting clinical outcome

of epidermal growth factor receptortargeted therapy in

metastatic colorectal cancer. JNCI.

Simon, A. (2010). Cosmic: mining complete cancer

genomes in the catalogue of somatic mutations in can-

cer. Nucleic Acids Research.

Valdes, C. (2013). Characteristics of cross-hybridization

and cross-alignment of expression in pseudo-

xenograft samples by rna-seq and microarrays.

Journal of Clinical Bioinformatics.

Walther, A. (2000). Genetic prognostic and predictive

markers in colorectal cancer. Nature Reviews.Cancer.

BIOINFORMATICS2014-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

148