Food Control 91 (2018) 58e67

Contents lists available at ScienceDirect

Food Control
journal homepage: www.elsevier.com/locate/foodcont

Application of next generation semiconductor based sequencing for

species identification and analysis of within-species mitotypes useful
for authentication of meat derived products
Anisa Ribani a, 1, Giuseppina Schiavo a, 1, Valerio Joe Utzeri a, Francesca Bertolini a, b,
Claudia Geraci a, Samuele Bovo a, Luca Fontanesi a, *
a
Department of Agricultural and Food Sciences (DISTAL), Division of Animal Sciences, University of Bologna, Viale Fanin 46, 40127, Bologna, Italy
b
Department of Animal Science, Iowa State University, 2255 Kildee Hall, 50011, Ames, IA, USA

a r t i c l e i n f o a b s t r a c t

Article history: In this study, we tested the Ion Torrent next generation semiconductor based sequencing technology for
Received 7 January 2018 meat species identification in several highly processed and complex meat products and meat derived
Received in revised form broths (a do € ner kebab, a beef/pork pate
, a meat based filling of tortellini, one instantaneous granular
26 February 2018
preparation of broth stock made by meat and two ready to use meat broths from different producers).
Accepted 22 March 2018
Available online 24 March 2018
The detection protocol included the sequencing of targeted mitochondrial DNA (mtDNA) regions
amplified with universal primer pairs and a bioinformatic pipeline designed to interpret sequencing
results. Six libraries were sequenced producing a total of 1,363,351 filtered reads. Data mining detected
Keywords:
Authenticity
expected and unexpected meat species in the analysed products. Pork was identified in the kebab and
Meat species identification Bubalus bubalis DNA was identified in the beef/pork pate . For products for which the precise meat species
Mitotype ingredient information was not available, it was possible to obtain it. Human contamination based on
mtDNA human detected reads could be useful to evaluate the hygienic level of highly processed products.
NGS Mitochondrial haplotypes (mitotypes) were identified for several mtDNA-species combinations
providing another level of information useful for the authentication of meat derived products. This work
defined a methodological framework to establish assays using this sequencing platform for routine
species identification in complex and highly processed food.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction
Purposely complete or partial substitutions of species in meat
products driven by economic advantages (i.e. frauds) and accidental
and uncontrollable mislabelling or mixings of meat from different
species (derived by errors occurring in the production plants or
over different steps of the production chains), raise many concerns.
Their implications are related to food safety and security, product
marketing and consumers’ acceptability, lifestyle, religious and
ethical issues (Ballin, 2010; Fontanesi, 2009, 2017). Therefore, the
identification of the species of origin of the products is becoming a
very relevant aspect in meat science. Many efforts have been made
to develop analytical methods and assays able to identify the
* Corresponding author.
E-mail address: luca.fontanesi@unibo.it (L. Fontanesi).
1
These authors contributed equally to this work.
species of origin in meat based and derived products, with appli-
cations also in other fields, including forensics (Fontanesi, 2017).
The identification of the species of origin in meat preparations is
mainly based on the differences that can be detected at the level of
biological components. Species-specific information useful for an-
imal species discrimination can be retrieved from proteins or
peptides, from some metabolites and from DNA (Fontanesi, 2017).
DNA has been extensively used for species identification in
different food products because i) it can be easily extracted from
most biological matrices, even after cooking or other physical
treatments due to its general high stability, and ii) it is possible to
capture sequence differences using PCR based assays. These
methods usually include an amplification step of targeted DNA
containing species-specific sequences and the analysis of the
amplified region with different approaches that have in common,
as final step, the separation of DNA fragments by gel electropho-
resis (i.e.: species-specific PCR, based on the presence or absence of
amplification from targeted species; PCR-RFLP, based on the use of

https://doi.org/10.1016/j.foodcont.2018.03.034
0956-7135/© 2018 Elsevier Ltd. All rights reserved.
 A. Ribani et al. / Food Control 91 (2018) 58e67
restriction enzymes that cut DNA palindromes; PCR-fingerprinting
techniques, including PCR-RAPD, PCR-AFLP, PCR-SSCP, PCR-DGGE,
high resolution melting, which produce species-specific DNA pat-
terns using several methods; PCR-FINS, based on Sanger
sequencing of amplified products; Mafra, Ferreira, & Oliveira, 2008;
Fajardo, Gonza lez, Rojas, García, & Martín, 2010; Rodríguez-
Ramírez, Gonz alez-Cordova, & Vallejo-Cordoba, 2011; Fontanesi,
2017). The targeted DNA assays usually amplify informative mito-
chondrial DNA (mtDNA) fragments [i.e. 12S, 16S, cytochrome b and
cytochrome c oxidase I (COI) genes or the D-loop region] which are
flanked by conserved sequences across many species that are used
for the design of universal PCR primers, with intervening regions
that contain species-specific sequences used for species identifi-
cation. Species-specific PCR are designed as end-point PCR or
quantitative real time PCR without the need of gel-electrophoresis
separation of the amplicons (Fontanesi, 2017). All these methods
can usually detect one or few species at the time. Therefore, the
choice of the assays is quite difficult when more than one species is
expected or when there is no at priori information on the presence
of species in mixed and processed meat or in derived products like
meat broth. To overcome the poor multiplexing detection potential
of these methods, microarray detection systems have been
designed to discriminate on the same chip, at the same time,
mtDNA amplified fragments belonging to several species
(Chisholm, Conyers, & Hird, 2008; Cottenet et al., 2016; Iwobi,
Huber, Hauner, Miller, & Busch, 2011; Peter, Brünen-Nieweler,
Cammann, & Bo €rchers, 2004; Teletchea, Bernillon, Duffraisse,
Laudet, & Ha €nni, 2008; Wang, Zhu, Chen, Xu, & Zhou, 2015).
These systems are more informative than simple PCR based assays
but are still limited by the construction bias of the tool that can
detect only species for which probes are already included on the
chip.
Next generation sequencing (NGS) technologies have changed
the way in which DNA can be analysed. These technologies have
increased of several orders of magnitude the sequencing
throughput. Their advantage in the context of species identification
relies on the possibility to combine in a single step the generation of
species-specific information from the produced short sequence
reads with limitless multiplexing potential and the possible relative
quantification of the detected species by counting reads matching
the same target specific sequence (even if, in many cases, this
estimation might depend by the experimental design and technical
procedures). The large potential and flexibility of NGS is obtained
through bioinformatic analysis of the produced sequence data
(Bertolini et al., 2015; Pabinger et al., 2014). Among the commer-
cially available benchtop NGS platforms that can be used for species
identification, it is possible to mention Roche 454, Illumina and Ion
Torrent technologies (Bertolini et al., 2015; Goodwin, McPherson, &
McCombie, 2016; Tillmar, Dell'Amico, Welander, & Holmlund,
2013). Ion Torrent technology is based on the detection of small
pH modifications that occur during the sequencing phase and that
are captured by semiconductor chips accommodating millions of
reaction micro-wells that produce sequence reads (Rothberg et al.,
2011). A few reports have tested the Ion Torrent Personal Genome
Machine (PGM) sequencing platform for species identification in
food products (i.e. Carvalho, Palhares, Drummond, & Gadanho,
2017; Giusti, Armani, & Sotelo, 2017; Mun ~ oz-Colmenero,
Martínez, Roca, & Garcia-Vazquez, 2017). We recently evaluated
this technology for species detection in DNA mixtures of meat
species and for species identification in dairy products (Bertolini
et al., 2015; Ribani et al., 2018).
In this study, we tested the usefulness of this NGS technology for
animal species identification in several highly processed and
complex meat products and meat derived broths. The study defined
a methodological framework to establish assays based on this NGS
59
platform for species identification in complex meat derived
matrices. The detection protocol included the sequencing of tar-
geted mtDNA regions amplified with three different universal
primer pairs and a bioinformatic pipeline designed to interpret
sequencing results. In addition, this study described the possibility
to identify within-species mitotypes, adding another level of in-
formation useful for the authentication of food products derived
from meat species.
2. Materials and methods
2.1. Analysed products
Six different products were investigated in this study: one
commercial do €ner kebab preparation; one beef/pork pate ; one
meat based filling of tortellini; one instantaneous granular prepa-
ration of broth stock made by meat; two ready to use meat-based
broths from different producers. More details on these products,
including their declared ingredients and origin, are reported in
Table 1. These products represent highly processed meat prepara-
tions derived or potentially derived from different species.
2.2. DNA isolation
A total of 0.5 g or 0.5 ml for each of these products was used for
DNA extraction. DNA was isolated using the Wizard® Genomic DNA
Purification kit (Promega Corporation, Madison, WI, USA) following
the manufacturer's instructions for DNA isolation from cell cultures
and animal tissues. The quality of the extracted DNA was assessed
using the Nanophotometer P-330 instrument (Implen GmbH,
München, Germany) and by visual inspection on 1% agarose gel
electrophoresis in TBE 1X buffer after staining with 1X GelRed
Nucleic Acid Gel Stain (Biotium Inc., Hayward, CA, USA).
2.3. PCR
Amplicons from the selected products were obtained using
three universal primer pairs that have been already shown to
successfully amplify many different vertebrate species (Table 2).
These primers were designed in conserved regions of the 12S and
16S mitochondrial rRNA genes (Bertolini et al., 2015; Karlsson &
Holmlund, 2007; Kitano, Umetsu, Tian, & Osawa, 2007; Ribani
et al., 2018). PCR amplifications were obtained using a 2720 ther-
mal cycler (Life Technologies, Carlsbad, CA, USA) in a total reaction
volume of 20 mL. PCR mix components were: 2X of the Kapa HiFi
HotStart ReadyMix PCR kit (Kapa Biosystems, Boston, Massachu-
setts, USA), 10 pmol of each primer and 50 ng of template DNA. The
amplification profile was the following: an initial denaturation step
of 3 min at 98
C; then 35 cycles of 20 s at 98
C, 15 s at the primer
pair specific annealing temperature (Table 2), 30 s at 72
C; and a
final extension step of 1 min at 72
C. Results of the PCR amplifi-
cations were evaluated visualizing the obtained fragments by 2.5%
agarose gel electrophoresis in TBE 1X buffer after staining with 1X
GelRed Nucleic Acid Gel Stain (Biotium Inc., Hayward, CA, USA).
Quantity and quality of the amplified fragments was estimated
using a Nanophotometer P-330 instrument (Implen GmbH).
2.4. Ion Torrent sequencing
The amplified fragments were purified with ExoSAP-IT® (USB
Corporation, Cleveland, Ohio, USA). Six libraries (one for each
product) were prepared for sequencing with the Ion Torrent PGM
(Thermo Fisher Scientific Inc.). Each library was based on amplicons
of the three primer pairs obtained for the same product. The li-
braries included end-repaired and barcoded amplicons (200 ng of
60 A. Ribani et al. / Food Control 91 (2018) 58e67

Table 1
List of products analysed in this study and their declared ingredients.

Library Product Labelled meat ingredients Declared origin Other ingredients

1 €ner kebab preparation

Do
2 
Beef/pork pate
3 Meat based filling of tortellini
4 Instantaneous granular
preparation of broth stock
made by meat
5 Ready to use liquid meat based
broth (producer A)
6 Ready to use liquid meat based
broth (producer B)
Unknown Unknown Unknown
Beef (25%), pork (20%), pig lard Produced and Lactose and other components (rice flour, wheat fiber,
packed in Italy guar gum, monosodium glutamate, sodium nitrate)
Pork (derived from meat, dry cured Produced and Grana Padano cheese, milk serum and other
ham and mortadella), small fraction packed in Italy components (salt, sugar, wheat flour, sodium citrate,
of beef spices)
Extract of meat (unspecified Produced and Lactose and other components (salt, monosodium
species) packed in Italy glutamate, sulphur dioxide, soybean, extract of
vegetables, onion, sedan, garlic, tomato, anti-oxidants)
Extract of unspecified meat (0.07%), Produced and Eggs, milk, vegetables (leek, tomato, carrot), yeast
extracted of beef (0.04%), extracted packed in Italy extract, salt, natural flavor of sedan, spices, extra-virgin
of chicken meat (0.04%) olive-oil
Extract of unspecified meat (0.2%), Produced and Vegetables (onion, sedan, tomato, carrot), yeast extract,
extract of beef (0.04%), extract of packed in Italy salt, natural flavor
chicken meat (0.04%)

Table 2
Universal PCR primers pairs used for DNA amplification of the 12S and 16S targeted mtDNA regions.

Primer pair name/mtDNA region PCR primers (50 -30 ): forward and reverse Annealing T (
C) Size of the amplified fragment (bp)a References

12S_Ki CCCAAACTGGGATTAGATACCC 59 215e219 Kitano et al. (2007)

GTTTGCTGAAGATGGCGGTA
16S_KH GACGAGAAGACCCTATGGAGC 59 112e114 Karlsson and Holmlund (2007)
TCCGAGGTCGCCCCAACC
16S_Ki GCCTGTTTACCAAAAACATCAC 62 244e245 Kitano et al. (2007)
CTCCATAGGGTCTTCTCGTCTT
a
The size of the amplified regions is different in the considered species. The range is reported.

amplified DNA) produced with the Ion Xpress™ Plus Fragment Li-
brary and Ion Xpress™ Barcode Adapters 1e16 kits (Thermo Fisher
Scientific Inc.). The Ion Library Quantitation kit (Thermo Fisher
Scientific Inc.) was used for the qPCR analysis carried out with the
StepOnePlus™ Real-Time PCR System (Thermo Fisher Scientific
Inc.) for the quantification of the prepared libraries. Then, the six
barcoded and quantified libraries were pooled together at the same
concentration and submitted to clonal amplification by emulsion
PCR with the Ion PGM™ Hi-Q™ OT2 kit (Thermo Fisher Scientific
Inc.). Sequencing was obtained in an Ion Torrent PGM (Thermo
Fisher Scientific Inc.) using the Ion PGM™ Hi-Q™ Sequencing kit
and an Ion 318 v2 chip (Thermo Fisher Scientific Inc.) following the
manufacturer's instructions.
2.5. Next generation sequencing data analyses
2.5.1. Data processing
Data were processed as previously described (Bertolini et al.,
2014). Briefly, reads produced from the Ion Torrent sequencing
run were first processed with the Torrent Suite v.4.6 on the Ion
Torrent Server (Thermo Fisher Scientific Inc.). Then, obtained FASTQ
files were quality checked using fastqc tool (http://www.
bioinformatics.bbsrc.ac.uk/projects/fastqc/). Pre-processed reads
were subsequently attributed to the six libraries according to their
barcodes and were then filtered by eliminating polyclonal and low-
quality sequences (Q < 20). Then, sequences of the adapters, of the
barcodes and low quality 30 -ends were trimmed from the filtered
reads. After these automatic processes, reads were then trimmed
from the PCR primer sequences at 50 and 30 end using the trim
function of HOMER. Retained reads were at least 2/3 in length of the
expected size of the amplicons produced with the three primer
pairs.
Next generation sequencing data were deposited in the EMBL-
EBI European Nucleotide Archive (ENA) with the project accession
number PRJEB25178.
2.5.2. Species identification
Trimmed reads were aligned to pre-built reference sequences of
the three amplified mtDNA regions of a customized database of
vertebrate species that was constructed for this study. Sequences of
more than 500 species were included in the database, according to
their potential relevance as ingredients or possible accidental
contaminants of meat products (all farm animals, the most com-
mon mammals, birds, reptiles and amphibians, including game
species and endangered species). Alignments were based on bwa
v.0.7.11 (Li et al., 2009) using the aln algorithm with default pa-
rameters or using a quality score of the alignment Qm > 30 (Li &
Durbin, 2009). Among the tested algorithms of previous works
(Bertolini et al., 2015; Ribani et al., 2018), aln performs better to
discriminate close homologous sequences (Li & Durbin, 2009).
Thus, aln was selected in this study considering the sequence
closeness of the analysed mtDNA regions of many meat species
(Ribani et al., 2018). Bam files were obtained using Samtools soft-
ware v.0.1.16 (Li et al., 2009). In addition, all reads were analysed
using BLASTþ v. 2.2.30þ (ftp://ftp.ncbi.nlm.nih.gov/blast/
executables/blastþ/2.2.30/) against all mtDNA vertebrate se-
quences available in GenBank (https://www.ncbi.nlm.nih.gov/
nucleotide/; November 15, 2017). Bioinformatic analyses were run
using Galaxy platform (Goecks, Nekrutenko, Taylor, & Team, 2010)
and considering matches with identity  99% over  90% of the
length of the sequenced read.
Other parameters useful to call reads were obtained from the
experiment of Ribani et al. (2018) that was carried out in parallel
with the Ion Torrent run of the current study (see below). Reads
were attributed to one species if: i) at least two nucleotides
differentiated it from another species (over the considered read
sequence); and ii) the proportion of reads assigned to a species was
higher than the product of the error rate of the Ion Torrent
sequencing for the amplified fragments, calculated on the number
of nucleotide differentiating two closely related species. The
average error rate for the considered amplicons has been already
 A. Ribani et al. / Food Control 91 (2018) 58e67
defined for different species: 0.051 for the 16S_KH fragments; 0.022
for the 16S_Ki and 12S_Ki fragments (Ribani et al., 2018). In addi-
tion, as reported in our previous work (Ribani et al., 2018), we
adopted the threshold of 0.2% as the proportion of the number of
reads assigned to a species (over all reads of that amplified region
within the considered library) to call the presence of a species in
the analysed product. This conservative threshold was defined to
overcome the potential problems derived by laboratory accidental
DNA contamination of meat animal species and by spurious bio-
informatic algorithm assignment of reads to non-correct species
(Ribani et al., 2018). Therefore, this threshold identified the limit of
detection based on the number of reads assigned to a meat species.
Evaluation of human DNA accidental contamination in the lab-
oratory environment that run the Ion Torrent sequencing experi-
ment was already evaluated by Ribani et al. (2018). Briefly, the
experiment of Ribani et al. (2018), who sequenced amplicons
derived by artificially prepared DNA pools constructed using cattle,
buffalo, goat and sheep DNA, showed that the number of reads
matching the three human targeted sequences ranged from 0 to 4
over a total of 40,000e160,000 reads per library, providing an es-
timate of the proportion of reads derived by contaminating human
DNA in the laboratory environment. This evaluation is below the
threshold of 0.2% of reads assigned to a species to call the presence
of that species in the analysed products. Therefore, as a conserva-
tive approach, we used the same threshold of 0.2% to consider the
presence of contaminating human DNA of the investigated
products.
2.5.3. Identification of mitochondrial haplotypes
The identification of within-species mitochondrial haplotypes
(i.e. mitotypes) from the sequenced reads was obtained as follows.
First, the GenBank database was queried with the reference se-
quences for each species that were identified in the analysed
products (i.e. >0.2% of reads) using BLASTN to identify mitotypes
for the corresponding regions already deposited in the nucleotide
database and derived by other studies. Then, to overcome the
problem of sequencing errors in defining mitotypes (that in any
case had to be already represented in the GenBank database), only
mitotypes whose read proportion was higher than the sequencing
error rate for that specific fragment (Ribani et al., 2018), over all
reads assigned to that species, were considered. Then, within-
species mitotypes were identified using aln alignments with
scripts to automatically retrieve different mitotype sequence in-
formation (defined by BLASTN analysis as described above) on the
aligned reads. The EMPOP database, version 3.0, release 11 (https://
empop.online/; Parson & Dür, 2007), was queried to identify the
mtDNA derived bio-geographic origin of the human contaminating
DNA based on the identified mtDNA haplotypes.
3. Results
3.1. Ion Torrent reads
A total of 3,052,478 raw reads were obtained from the Ion
Table 3
Number of sequence reads produced from the six analysed libraries.
Library Product
1 Do€ner kebab preparation
2 
Beef/pork pate
3 Meat based filling of tortellini
4 Instantaneous granular preparation of broth stock made by meat
5 Ready to use liquid broth based on meat (producer A)
6 Ready to use liquid broth based on meat (producer B)
61
Torrent run. These reads were approximately equally distributed
among the six barcoded libraries (ranging from 15.4% for library
5e18.4% for library 6 (Table 3). Raw reads were filtered according to
the size and sequencing quality. After these processing steps, a total
of 1,363,351 reads (44.7%) remained for further analyses. Of the six
libraries that were analysed, library 1 retained only 32.5% of the
reads and library 4 retained 61.1% of reads. The filtered reads were
assigned to the three amplified fragments with the following pro-
portion: 426,924 (31.3%) were of the 12S_Ki fragment, 686,587
(50.3%) belonged to the 16S_Ki fragment and 250,840 (18.4%) were
from the 16S_KH region. The lower proportion of reads obtained for
the 16S_KH region over all filtered reads indirectly confirmed the
lower efficiency of amplification of this primer pair, compared to
the other two pairs for the amplification of mammalian DNA
(Bertolini et al., 2015).
3.2. Meat species identified in the analysed products
The number of reads of the three investigated mtDNA regions
assigned to different mammalian or avian species for each analysed
product is reported in Fig. 1. Results obtained using aln software
(with mapping quality Qm > 30 and without a mapping threshold)
and BLASTN analysis are included for comparison. Fig. 2 shows the
distribution (in percentage) of reads of the three mtDNA fragments
attributed to different species as determined by BLASTN. The dis-
tribution of reads assigned to species as obtained by aln software
(with mapping quality Qm > 30 and without a mapping threshold)
is reported in the Supplementary Material (Figure S1 and Figure S2,
respectively). Complete data are reported in Table S1 of the
Supplementary Material.
Analysis of the reads obtained from the do €ner kebab, for which
meat species ingredients was not known, indicated the presence of
meat from two main species (cattle and chicken) confirmed with
the three amplified mtDNA fragments (using all three bioinformatic
methods) even if with different estimated proportions. The 16S_KH
and 16S_Ki fragments showed a similar percentage of reads of the
two species, whereas reads of the 12S_Ki fragment were mainly
assigned to chicken. Unexpectedly, all three fragments and bio-
informatic analyses reported also the presence of pork. Reads from
pig were just above the 0.2% threshold for two of the amplified
regions (12S_Ki and 16S_KH) and were about 10% for the 16S_Ki
fragment. No other meat species showed at least two regions with
more than 0.2% of assigned reads.
Reads of the beef/pork pate  library indicated the presence of
three meat species: cattle (Bos taurus), buffalo (Bubalus bubalis) and
pig (Sus scrofa). The proportion of reads of the two Bovidae species
was similar for all three mtDNA regions and bioinformatic ap-
proaches (7e11% cattle; 32e44% buffalo), accounting, on the whole,
for about 41e55% of all reads of this library. Close inspection of read
alignments to the reference buffalo and cattle sequences confirmed
that the bioinformatic pipelines assigned correctly the sequences to
these two species. Buffalo meat was not indicted in the product
label as a component of this pate . Reads assigned to Sus scrofa were
about 40e55% over the total (except for the BLASTN analysis of the
Total no. of Filtered 12S_Ki Filtered 16S_Ki Filtered 16S_KH No. of total
raw reads reads reads reads filtered reads
501397 309493 21951 173305 186190
469847 141934 40305 242542 257631
518667 113503 46299 145167 452820
520911 140464 112868 272425 358069
481005 131017 82186 181462 394665
560651 94784 47689 222685 438409
62 A. Ribani et al. / Food Control 91 (2018) 58e67

Fig. 1. Number of reads (y axis) obtained from the six libraries and for the three analysed mtDNA regions (12S_Ki, 16S_KH and 16S_Ki) that were assigned to different species by aln
(using Qm30 and default parameters; blue and orange bars, respectively) and BLASTN (grey bars; x axis). Species reported are only those for which some reads were assigned using
at least one of the methods (mtDNA region and/or bioinformatic approach) applied in this study. Species are indicated with the following acronyms: BT, Bos taurus (cattle); BB,
Bubalus bubalis (buffalo); GG, Gallus gallus (chicken); HS, Homo sapiens (humans); LE, Lepus europaeus (hare); OC, Oryctolagus cuniculus (rabbit); SS, Sus scrofa (pig). (For inter-
pretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
 A. Ribani et al. / Food Control 91 (2018) 58e67
Fig. 2. Proportion of reads of the three analysed mtDNA fragments (12S_Ki, 16S_KH and 16S_Ki), obtained from the six libraries, assigned to different species with BLASTN.
16S_KH that were about 23% of all filtered reads).
Reads produced from meat-based filling of tortellini for the
three amplified fragments indicated that this product contained
meat from pig only. The small fraction of the declared beef or dairy
cattle cheese components of the tortellini meat mixture (Table 1)
did not contribute to the detected species in this product. All three
mtDNA fragments did not identify Bos taurus as reads assigned to
this species did not trespass the 0.2% threshold (reads assigned to
this species were about 0.1%). It could be possible that the pre-
defined limit of detection could be too stringent for this product.
Meat species (not precisely declared in the label; Table 1)
identified in the instantaneous granular preparation of broth stock
made by meat were: 1) pig, that accounted for about 2/3 of all reads
(for all three amplified regions and bioinformatic methods); 2)
cattle (5e9% of reads for all three targeted mtDNA fragments and
bioinformatic analyses); 3) rabbit (about 1e3% of reads for all three
amplicons and bioinformatic approaches); and 4) chicken (about
2% for two out of three mtDNA regions).
Bos taurus accounted for the largest number of reads in library 5
(derived from a ready to use liquid meat-based broth - from pro-
ducer A). Aln and BLASTN analyses reported about 84e98% of reads
obtained from all three primer pairs assigned to this species. Aln
software using Qm > 30 as matching criteria reported, for all three
analysed regions, about 20% of reads assigned to this species. This
reduction of percentage might be due to the performances of the
alignment algorithm that excludes from all mapped reads the
alignments below the Qm30 threshold (Ribani et al., 2018). Rabbit
reads were identified in two of the three amplified regions (12S_Ki
and 16S_Ki) over the 0.2% level. Aln Qm30 for the rabbit 12S_Ki
fragment reduced the number of reads probably caused by poor
alignment quality due to competitive mapping with hare se-
quences. This might be deduced from the results of BLASTN that, in
turn, cleaned the situation and did not assign any read to Lepus spp.
This liquid broth was expected to be derived from pig meat (0.04%,
according to the indication of the product label; Table 1) but the
results we obtained did not support the presence of this meat
component at the adopted stringency in terms of number of reads.
A less stringent threshold (i.e. 0.1%) would have identified the
presence of this species.
63
The meat species ingredients of the second ready to use liquid
broth based on meat (from producer B) was more complex than
that of the broth of producer A. Cattle (indicated in the label of this
product) was the meat species that accounted for the largest
number of reads for all three mtDNA amplicons (from about 30 to
65%, depending by the amplified fragment and applied bio-
informatic method). Pig accounted for about 4e14% of reads and
rabbit reads were from about 0.2% to 1.4% over the total of reads.
Chicken (mentioned in the label of this product) accounted for
about 5% of reads obtained for the 12S_Ki and 16S_KH regions (two
out of three fragments, as the 16S_Ki reported a very low number of
reads).
3.3. Identification of human DNA contamination
The universal primers used in this study also amplify with high
efficiency the human corresponding mtDNA regions (Bertolini
et al., 2015; Karlsson & Holmlund, 2007; Kitano et al., 2007).
Therefore, considering that in the parallel experiment (that shared
all steps at the same time, from DNA extraction to library prepa-
ration and PGM run) we were able to exclude the accidental human
contamination of the laboratory environment (Ribani et al., 2018),
we used the number of mapped reads on the human mtDNA se-
quences as a parameter useful to evaluate the level of human
contamination in the analysed products. The minimum number of
reads to identify human contamination was the same used to
declare the presence of meat species (i.e. >0.2%) for at least two out
of three amplified regions.
Among the investigated products, the beef/pork pate  and the
meat based filling of tortellini did not shown any human contam-
ination. For these two products, the number of reads assigned to
human mtDNA sequences was lower than the adopted background
level (i.e. 0.2%), suggesting absence of contamination (or limited
contamination) of human origin on these meat preparations. The
do€ner kebab preparation reported human reads over the indicated
threshold for two out of three mtDNA fragments, i.e. 16S_KH (just
above the threshold) and 16S_Ki (with high proportion: >30%). All
three meat derived broths showed human contamination
confirmed by all three mtDNA regions and bioinformatic methods.
64 A. Ribani et al. / Food Control 91 (2018) 58e67

About 2e4% of reads obtained from library 4 (derived by the
instantaneous granular preparation of broth stock made by meat)
were mapped on human mtDNA regions. Human reads were about
1e2% (with 8e9% for the 16S_Ki using the aln software without
quality filtering or BLASTN analysis) in library 5, obtained from the
ready to use liquid broth of producer A). The most contaminated
product appeared the ready to use liquid broth of producer B (li-
brary 6). Reads assigned to human mtDNA were from about 17 to
56% (depending by the primer pairs and bioinformatic analysis).
3.4. Identification of species-specific mitotypes from Ion Torrent
reads
Next generation sequencing data from mtDNA fragments could
identify within species mitochondrial haplotypes (i.e. mitotypes).
To overcome the problem derived by sequencing errors, we
considered reliable mitotypes only those already reported in Gen-
Bank database and that their proportion among the sequenced
reads was higher than the sequencing error rate defined for that
specific fragment (Ribani et al., 2018). BLASTN analysis carried out
against GenBank deposited sequences (derived by other studies)
with the reference sequences of the three mtDNA regions (for all
species, for which we declared their presence in the investigated
products), identified a few mitotypes for some of the mDNA frag-
ment/species combinations (Table 4).
Among the possible seven porcine mitotypes of the 16_Ki re-
gion, only two (the haplotypes represented by the highest number
of deposited entries: 7 and 47, respectively) were identified in all
analysed products (including the broth from library 5, for which it
was not possible to declare the presence of pig meat according to
the 0.2% threshold; Table 4). Their proportions across samples was
different suggesting a different origin or contribution of different
pigs with these two mtDNA haplotypes to the analysed products.
This different proportion of porcine mtDNA mitotypes in the ana-
lysed products would also exclude cross contamination among the
analysed products.
Mining GenBank mtDNA buffalo sequences, three possible
mitotypes were identified in the 12S_Ki region (represented by 69,
45 and 12 sequence entries, respectively) and two possible
mitotypes were identified in the 16S_Ki region (represented by 114
and 37 entries, respectively; Table 4). The 12S_Ki region mitotype
that was present in 12 GenBank entries and the 16S_Ki mitotype
observed in 37 GenBank entries were the sequences that we
detected in the pate meat product.
Three rabbit mitotypes were identified in GenBank for the
12S_Ki mtDNA region (reported in three, one and one entries,
respectively; Table 4). Reads derived from the three meat broths
(the granular preparation of broth stock made by meat and the
ready to use liquid broths based on meat from the two producers),
were mainly assigned to the most frequent form (mitotype 1) and
to mitotype 2 already present in GenBank.
Two possible human 16S_Ki mitotypes were retrieved from
GenBank (Table 4). These two mitotypes were identified in all
sequenced products for which the human contamination was
>0.2% of reads, even if with different proportions. The presence of
different proportions of these two mitotypes in the analysed
products indirectly excludes that the human contamination in the
sequenced food matrices derives from the laboratory environment.
In that case, we would have expected uniform human mtDNA
haplotype distribution across samples, considering also that lab
personnel carried only one of these haplotype (data not shown).
Even if the covered mtDNA fragment region has a limited infor-
matively, using EMPOP database we obtained some information on
the probable ethnicity of the human DNA contamination. Haplo-
types 1 and 2 were present in 65.8% and 29.6% of human mtDNA
fragments deposited in this database with high frequency in Eu-
ropean and American populations (haplotype 1: 61.1% and 100% of
the mtDNAs from these populations) or African and European
populations (haplotype 2: 71.4% and 34.9% of the mtDNAs from
these populations).
4. Discussion
A few NGS approaches have been recently applied for species
identification or for the detection of other signatures useful for the
authentication of food products, demonstrating the powerful po-
tential of these DNA sequencing methods. For example, NGS tech-
nologies have been used for species identification in complex or

Table 4
Summary of mitotypes identified for different species and mtDNA fragments based on entries already reported in GenBank database.

Species Fragment Mitotype IDa Accession numberb Occurrence in databasec No. of readsd

Library 1 Library 2 Library 3 Library 4 Library 5 Library 6

Sus scrofa 16S_Ki 1 KX982632 7 2404 28973 56112 57458 e 7164

2 EF545576 1 e e e e e e
3 MF183225 47 3532 26403 57309 83563 e 11802
4 AF304200 1 e e e e e e
5 KJ193222 1 e e e e e e
6 AY574047 1 e e e e e e
7 DQ274110 1 e e e e e e
Oryctolagus cuniculus 12S_Ki 1 AY292691 3 e e e 1916 51 117
2 FM164771 1 e e e 39 36 56
3 FM164772 1 e e e e e e
Bubalus bubalis 12S_Ki 1 KX758295 12 e 26892 e e e e
2 KX758396 14 e e e e e e
3 KX758402 69 e e e e e e
16S_Ki 1 KX758402 114 e e e e e e
2 KX758295 37 e 35100 e e e e
Homo sapiens 16S_Ki 1 KM101984 169 10697 e e 577 528 76769
2 AF382008 76 10330 e e 5736 1426 2513
a
Mitotype progressive number (ID).
b
GenBank accession number of sequences taken as example for the different mitotypes.
c
Number of entries already deposited in GenBank having the same sequence of the indicated mitotypes.
d
Number of reads identified in the different libraries attributed to the corresponding mitotypes. The number of reads reported in the table is slightly different from the total
number of mapped reads of the same species due to the capturing information of the bioinformatic scripts that were used to discriminate reads matching different mitotypes.
The numbers are reported only for the species for which reads trespassed the 0.2% threshold applied to declare its presence in the evaluated library.
 A. Ribani et al. / Food Control 91 (2018) 58e67
highly processed seafood products (Abbadi et al., 2017; De Battisti
et al., 2014; Giusti et al., 2017; Giusti et al., 2017; Park et al.,
2012), for the characterization of food species in candies (Mun ~ oz-
Colmenero et al., 2017), for species detection in dairy products
(Ribani et al., 2018) and for the determination of the botanical
origin of honey (Utzeri et al., 2018). Species identification in meat
products has been approached by Ripp et al. (2014) using whole
DNA sequencing of sausage DNA using Illumina sequencing plat-
forms. We recently evaluated a next generation semiconductor-
based sequencing platform for the detection of meat species in
DNA mixtures (Bertolini et al., 2015). In this study, we applied this
sequencing technology for species identification in complex and
highly processed meat matrices and derived products using a tar-
geted mtDNA approach based on different fragments amplified
with three universal primer pairs.
These methods can identify, at the same time, all animal species
that are present in the analysed products, including unexpected
species that were not suspected to contribute to or to contaminate
the products. Therefore, NGS methods can be applied to complex
and highly processed food products that underwent several phys-
ical treatments and preparation steps. In addition, mining
sequencing data with appropriate bioinformatic pipelines and us-
ing customized databases, it is possible to extract other information
that is contained in the obtained sequences that might be useful to
trace, without any extra efforts or costs, the food origin by detecting
intra-species variability (that in our study was investigated at the
mtDNA level). This information could be also useful for forensic
investigations.
We already evaluated several technical aspects, parameters and
bioinformatic algorithms to produce and analyse sequencing reads
obtained with the Ion Torrent platform and assign them to refer-
ence animal species (Bertolini et al., 2015; Ribani et al., 2018). In this
study, we took advantage from what we previously determined as
sequencing error rate of the three mtDNA fragments tested in this
study, background noise (including the level of laboratory back-
ground DNA that run the experiment) and limit of detection
(considered using the percentage of reads assigned to a species
with the 0.2% threshold) determined by the combined sensitivity of
the sequencing technology and bioinformatic methods to identify a
species from the produced reads (Bertolini et al., 2015; Ribani et al.,
2018). This conservative threshold could be re-evaluated according
to the results that we expected on a few products for which the 0.1%
threshold could have detect declared minor components in li-
braries 3 and 5. The bioinformatic pipeline relied on the results of
the aln algorithm that was already shown to perform better for
species identification discriminating closely related sequences
(Ribani et al., 2018). The introduction of BLASTN analyses against
the GenBank vertebrate mtDNA sequence database further
extended the possibility to identify many different species. In this
study we compared the results of this algorithm with the outputs of
BLASTN to add another computational method and validate what
obtained with aln software whose sensitivity might discard a quite
large number of reads (reducing the number of useful information)
if stringent alignment quality criteria are applied. This is the case
that we observed in library 5 (derived from a ready to use liquid
meat based broth, producer A) where reads attributed to Bos taurus
dropped drastically if aln Qm30 was considered. This might be due
to the possibility that some true cattle reads could be also aligned
(with similar score) to reference sequences of other close species
having high sequence identity, preventing the use of these reads for
unique species identification. We also used an internal validation
approach to declare the presence of a species based on the same
results obtained by at least two out three concordant mtDNA re-
gions amplified and sequenced for each product. These criteria
made it possible to overcome the problems derived by un-
65
supervised spurious bioinformatic alignments of the reads, labo-
ratory background DNA contamination and might consider the
need of an internal confirmation of the results. On the other hand,
the use of at least two and not all three fragments to call a species
might reduce the false negative results derived by PCR failure or
low efficiency in the amplifications for one of the three regions. It is
well known that PCR results and efficiency of some PCR primers
could be affected by inhibitors derived by complex food matrices
that cannot be completely eliminated during the DNA extraction
steps (i.e. Demeke & Jenkins, 2010; Hellberg, Kawalek, Van, Shen, &
Williams-Hill, 2014).
In addition to the two out of three fragment criteria, we added,
as additional level needed for declaring the presence of a species,
the confirmation of the results by the BLASTN. BLASTN was able to
eliminate some spurious results as in the case of the hare DNA. Its
presence in library 5 (derived from a ready to use liquid meat-based
broth, producer A) identified with the aln software was not
confirmed and reads reattributed to the rabbit species.
Unexpected meat species were identified in some products that
we analysed. For example, the do €ner kebab for which we did not
know the declared meat species ingredients (and thus it was used
as a blind test in our study) reported the presence of pork. Ac-
cording to the typical origin and characteristic of this product, pork
should not be present. It seems however that contaminating pork is
quite common in these products, including western halal
commercialized preparations (Soares, Amaral, Oliveira, & Mafra,
2013).
The beef/pork pate  reported the presence of buffalo (Bubalus
bubalis) DNA. This result was not expected as the substitution of
beef with buffalo meat is not a common fraud. The identification of
Bubalus bubalis species is not easy as most end-point or quantitative
real time PCR methods designed to identify beef cannot discrimi-
nate closely related species and thus the presence of buffalo DNA is
not usually recognized. Buffalo meat could be considered a by-
product of the buffalo mozzarella production chain and this
might be one possible explanation of the presence of buffalo meat
in highly processed products. Further studies are needed to eval-
uate the frequency of buffalo meat in highly processed beef labelled
products and the origin of this possible fraud or accidental mixture
of Bovidae species.
All three broth preparations analysed in this study reported
general indications on the presence of undefined meat species. Two
of them declared in the label, at least in part, the animal species
from which they were produced. We identified that the instanta-
neous granular preparation of broth stock made by meat was
derived by four different meat species (cattle, pig, rabbit and
chicken). The label of the ready to use liquid meat based broth from
producer A reported the presence of the pig species at a very low
level. However, read analysis did not identify this species, probably
due to the complete absence of material from the pig despite the
information present in the label. The other ready to use liquid meat
based broth (from producer B) resulted to be derived from four
species (cattle, pig, rabbit and chicken). The rabbit was not declared
in the label of this product even if meat extract of unspecified meat
species was declared (Table 1).
Analysis of reads produced from the six libraries did not identify
the presence of horse DNA or the presence of any other species than
those described (for example we did not identified the presence of
sheep or goat DNA) suggesting that, despite the analysed products
could be potentially derived from different species, the results
indicted that only a limited number of species were present. A more
extensive study on many other products should be carried out to
have a precise evaluation of the frauds and accidental contamina-
tions of undeclared species in meat derived products.
Another interesting result was the presence of human DNA
66 A. Ribani et al. / Food Control 91 (2018) 58e67
contamination in four out of six products. This parameter could be
related to the hygienic level over the production chain or of the
processing plants (Ribani et al., 2018). It could be also affected by
the degree of human manipulation of the products. Mun ~ oz-
Colmenero et al. (2017) reported a quite high level of human DNA
reads from an Ion Torrent NGS analysis designed for species iden-
tification in candies. They interpreted this result as derived by
external contamination inadvertently acquired during or after
candy processing and transferred later to the products resulting in
less degraded DNA than true ingredients, so constituting a better
template for PCR analyses of the targeted mtDNA products (Mun ~ oz-
Colmenero et al., 2017). The do €ner kebab and all three broths
showed human contamination. In particular, one ready to use broth
(from producer B) reported a worrying level of human contami-
nation confirmed by all three mtDNA regions. In general, the high
proportion of human reads in the liquid broths compared to other
meat products could be due to the fact that the level of meat species
DNA in the liquid preparations might be lower than that in solid
products. Therefore, this diluted presence of meat species DNA
could have some effects on the amplification efficiency and thus on
the relative proportion of human reads in these libraries.
It is however worth to mention that species identification with
the NGS approach evaluated in our study is not quantitative. The
number of reads attributed to different species roughly respects the
proportions derived by the sequenced DNA (Ribani et al., 2018).
This proportion might depend by the amplified fragment under its
specific analytical conditions, that would reduce the stability of the
results (in terms of consistency of the proportion of reads) across
PCR products. Several aspects, including different DNA quality and
quantity and different amplification efficiency in different species,
might introduce some biases preventing, in some cases, precise
quantitative measurements. For example, from the data reported
by Bertolini et al. (2015) it seems that the amplification and
sequencing of the three investigated mtDNA regions against human
DNA is more efficient than that on DNA of other livestock species.
Other studies are needed to better define these parameters and
then to use Ion Torrent sequencing data for quantitative analyses
(Bertolini et al., 2015).
By mining the obtained reads, it was possible to identify
different mitotypes for some species and fragment combinations.
Thus, the conservative bioinformatic approach that we adopted
(based on sequences already deposited in GenBank) was able to
identify within-species variability that added another level of in-
formation that could have interesting applications for the authen-
tication of meat products. Pig reads of the 16S_Ki fragment could be
assigned to two mitochondrial haplotypes. Their relative propor-
tion was different among the analysed, products providing a po-
tential signature of their origin and, indirectly, a proof that pig DNA
did not derive by environmental laboratory background. The same
could be deduced from the human DNA contamination reported in
four products that showed different proportions of the two human
mitotypes and for which we can exclude contamination from lab
personnel. Ethnicity of the human mtDNA sequences could be only
in part deduced as the 16S_Ki fragment is not very informative due
to its limited size. It could be possible to analyse other more
informative mtDNA regions and improve the usefulness of the
detection of human contaminating DNA also for forensic
applications.
5. Conclusions
This study applied the Ion Torrent semiconductor based
sequencing technology for species identification in highly pro-
cessed meat derived products. Thousands of reads were mined
with different bioinformatic approaches that reduced the
possibility of unsupervised spurious bioinformatic assignment of
reads to the wrong species. The tested method, that combined data
derived by three universal primer pairs with a customized bio-
informatic analysis to assure internal validation of the results,
might be useful for routine applications to identify expected and
unexpected species not only in meat products but also in other
animal derived products. An additional level of information could
be obtained by detecting mitotypes from the produced reads. The
empirical definitions of the limit of detection and of the level of
human DNA contamination might need further improvements.
Once these matters are set up, the described NGS approach might
constitute an analytical system that could become a reference
method for meat species identification in food products to detect
frauds and accidental contaminations. This NGS method could be
integrated to identify other components (i.e. plant species)
included in meat preparations.
Conflict of interest
The authors declare no conflicts of interest.
Acknowledgements
This work was developed as part of the GRIFFA research pro-
gram and was supported by University of Bologna 2017 RFO funds.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.foodcont.2018.03.034.
References
Abbadi, M., Marciano, S., Tosi, F., De Battisti, C., Panzarin, V., Arcangeli, G., et al.
(2017). Species identification of bivalve molluscs by pyrosequencing. Journal of
Agricultural and Food Chemistry, 97, 512e519.
Ballin, N. Z. (2010). Authentication of meat and meat products. Meat Science, 86,
577e587.
Bertolini, F., Ghionda, M. C., D'Alessandro, E., Geraci, C., Chiofalo, V., & Fontanesi, L.
(2015). A next generation semiconductor based sequencing approach for the
identification of meat species in DNA mixtures. PLoS One, 10, e0121701.
Bertolini, F., Schiavo, G., Scotti, E., Ribani, A., Martelli, P. L., Casadio, R., et al. (2014).
High-throughput SNP discovery in the rabbit (Oryctolagus cuniculus) genome by
next-generation semiconductor-based sequencing. Animal Genetics, 45,
304e307.
Carvalho, D. C., Palhares, R. M., Drummond, M. G., & Gadanho, M. (2017). Food
metagenomics: Next generation sequencing identifies species mixtures and
mislabeling within highly processed cod products. Food Control, 80, 183e186.
Chisholm, J., Conyers, C. M., & Hird, H. (2008). Species identification in food prod-
ucts using the bioMerieux FoodExpert-ID® system. European Food Research and
Technology, 228, 39e45.
Cottenet, G., Sonnard, V., Blancpain, C., Ho, H. Z., Leong, H. L., & Chuah, P. F. (2016).
A DNA macro-array to simultaneously identify 32 meat species in food samples.
Food Control, 67, 135e143.
De Battisti, C., Marciano, S., Magnabosco, C., Busato, S., Arcangeli, G., & Cattoli, G.
(2014). Pyrosequencing as a tool for rapid fish species identification and com-
mercial fraud detection. Journal of Agricultural and Food Chemistry, 62, 198e205.
Demeke, T., & Jenkins, G. R. (2010). Influence of DNA extraction methods, PCR in-
hibitors and quantification methods on real-time PCR assay of biotechnology-
derived traits. Analytical and Bioanalytical Chemistry, 396, 1977e1990.
Fajardo, V., Gonz alez, I., Rojas, M., García, T., & Martín, R. (2010). A review of current
PCR-based methodologies for the authentication of meats from game animal
species. Trends in Food Science & Technology, 21, 408e421.
Fontanesi, L. (2009). Genetic authentication and traceability of food products of
animal origin: New developments and perspectives. Italian Journal of Animal
Science, 8(Suppl. 2), 9e18.
Fontanesi, L. (2017). Meat authenticity and traceability. In F. Toldra  (Ed.), Lawrie's
meat science (Eight Edition, pp. 585e633). Duxford, UK: Woodhead Publishing,
Elsevier.
Giusti, A., Armani, A., & Sotelo, C. G. (2017). Advances in the analysis of complex
food matrices: Species identification in surimi-based products using Next
Generation Sequencing technologies. PLoS One, 12, e0185586.
Giusti, A., Tinacci, L., Sotelo, C. G., Marchetti, M., Guidi, A., Zheng, W., et al. (2017).
Seafood identification in multispecies products: Assessment of 16SrRNA, cytb,
and COI universal primers' efficiency as a preliminary analytical step for setting
 A. Ribani et al. / Food Control 91 (2018) 58e67
up metabarcoding next-generation sequencing techniques. Journal of Agricul-
tural and Food Chemistry, 65, 2902e2912.
Goecks, J., Nekrutenko, A., Taylor, J., & The Galaxy Team. (2010). Galaxy: A
comprehensive approach for supporting accessible, reproducible, and trans-
parent computational research in the life sciences. Genome Biology, 11, R86.
Goodwin, S., McPherson, J. D., & McCombie, W. R. (2016). Coming of age: Ten years
of next-generation sequencing technologies. Nature Reviews Genetics, 17,
333e351.
Hellberg, R. S., Kawalek, M. D., Van, K. T., Shen, Y., & Williams-Hill, D. M. (2014).
Comparison of DNA extraction and PCR setup methods for use in high-
throughput DNA barcoding of fish species. Food Analytical Methods, 7,
1950e1959.
Iwobi, A. N., Huber, I., Hauner, G., Miller, A., & Busch, U. (2011). Biochip technology
for the detection of animal species in meat products. Food Analytical Methods, 4,
389e398.
Karlsson, A. O., & Holmlund, G. (2007). Identification of mammal species using
species-specific DNA pyrosequencing. Forensic Science International, 73, 16e20.
Kitano, T., Umetsu, K., Tian, W., & Osawa, M. (2007). Two universal primer sets for
species identification among vertebrates. International Journal of Legal Medicine,
121, 423e427.
Li, H., & Durbin, R. (2009). Fast and accurate short read alignment with Burrows-
Wheeler transform. Bioinformatics, 25, 1754e1760.
Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., et al. (2009). 1000
genome project data processing subgroup. The sequence alignment/map (SAM)
format and SAMtools. Bioinformatics, 25, 2078e2079.
Mafra, I., Ferreira, I. M. P. L. V. O., & Oliveira, M. B. P. P. (2008). Food authentication by
PCR-based methods. European Food Research and Technology, 227, 649e665.
Mun ~ oz-Colmenero, M., Martínez, J. L., Roca, A., & Garcia-Vazquez, E. (2017). NGS
tools for traceability in candies as high processed food products: Ion Torrent
PGM versus conventional PCR-cloning. Food Chemistry, 214, 631e636.
Pabinger, S., Dander, A., Fischer, M., Snajder, R., Sperk, M., Efremova, M., et al. (2014).
A survey of tools for variant analysis of next-generation genome sequencing
data. Briefings in Bioinformatics, 15, 256e278.
Park, J. Y., Lee, S. Y., An, C. M., Kang, J. H., Kim, J. H., Chai, J. C., et al. (2012).
Comparative study between next generation sequencing technique and iden-
tification of microarray for species identification within blended food products.
67
Biochip Journal, 6, 354e361.
Parson, W., & Dür, A. (2007). EMPOPda forensic mtDNA database. Forensic Science
International: Genetics, 1, 88e92.
Peter, C., Brünen-Nieweler, C., Cammann, K., & Bo €rchers, T. (2004). Differentiation of
animal species in food by oligonucleotide microarray hybridization. European
Food Research and Technology, 219, 286e293.
Ribani, A., Schiavo, G., Utzeri, V. J., Bertolini, F., Geraci, C., Bovo, S., et al. (2018).
Application of next generation semiconductor based sequencing for species
identification in dairy products. Food Chemistry, 246, 90e98.
Ripp, F., Krombholz, C. F., Liu, Y., Weber, M., Scha€fer, A., Schmidt, B., et al. (2014). All-
food-seq (AFS): A quantifiable screen for species in biological samples by deep
DNA sequencing. BMC Genomics, 15, 639.
Rodríguez-Ramírez, R., Gonza lez-Cordova, A. F., & Vallejo-Cordoba, B. (2011). Re-
view: Authentication and traceability of foods from animal origin by poly-
merase chain reaction-based capillary electrophoresis. Analytica Chimica Acta,
685, 120e126.
Rothberg, J. M., Hinz, W., Rearick, T. M., Schultz, J., Mileski, W., Davey, M., et al.
(2011). An integrated semiconductor device enabling non-optical genome
sequencing. Nature, 475, 348e352.
Soares, S., Amaral, J. S., Oliveira, M. B. P., & Mafra, I. (2013). A SYBR Green real-time
PCR assay to detect and quantify pork meat in processed poultry meat products.
Meat Science, 94, 115e120.
Teletchea, F., Bernillon, J., Duffraisse, M., Laudet, V., & Ha €nni, C. (2008). Molecular
identification of vertebrate species by oligonucleotide microarray in food and
forensic samples. Journal of Applied Ecology, 45, 967e975.
Tillmar, A. O., Dell'Amico, B., Welander, J., & Holmlund, G. (2013). A universal
method for species identification of mammals utilizing next generation
sequencing for the analysis of DNA mixtures. PLoS One, 8, e83761.
Utzeri, V. J., Ribani, A., Schiavo, G., Bertolini, F., Bovo, S., & Fontanesi, L. (2018).
Application of next generation semiconductor based sequencing to detect the
botanical composition of monofloral, polyfloral and honeydew honey. Food
Control, 86, 342e349.
Wang, W., Zhu, Y., Chen, Y., Xu, X., & Zhou, G. (2015). Rapid visual detection of eight
meat species using optical thin-film biosensor chips. Journal of AOAC Interna-
tional, 98, 410e414.