Modelling molecular and inorganic data of Amanitaponderosa mushrooms using artificial neural networks

Catia Salvador • M. Rosario Martins •

Henrique Vicente • Jose Neves • Jose M. Arteiro •

A. Teresa Caldeira

Received: 3 January 2012 / Accepted: 9 July 2012 / Published online: 24 July 2012

� Springer Science+Business Media B.V. 2012

Abstract Wild edible mushrooms Amanita ponder-

osa Malencon and Heim are very appreciated in

gastronomy, with high export potential. This species

grows in some microclimates, namely in the southwest

of the Iberian Peninsula. The results obtained demon-

strate that A. ponderosa mushrooms showed different

inorganic composition according to their habitat and

the molecular data, obtained by M13-PCR, allowed to

distinguish the mushrooms at species level and to

differentiate the A. ponderosa strains according to

their location. Taking into account, on the one hand,

that the characterisation of different strains is essential

in further commercialisation and certification process

and, on the other hand, the molecular studies are quite

time consuming and an expensive process, the devel-

opment of formal models to predict the molecular

profile based on inorganic composition comes to be

something essential. In the present work, Artificial

Neural Networks (ANNs) were used to solve this

problem. The ANN selected to predict molecular

profile based on inorganic composition has a 6-7-14

topology. A good match between the observed and

predicted values was observed. The present findings

are wide potential application and both health and

economical benefits arise from this study.

Keywords Ectomycorrhizal macrofungi �Wild

edible mushrooms � M13-PCR � Inorganic

composition � Artificial intelligence based tools

Introduction

Amanita ponderosa (Malencon and Heim 1942) are

wild edible mushroom, growing in some mediterra-

nean microclimates, in the southwest of the Iberian

Peninsula, due to its mediterranean characteristics,

namely in Alentejo (Portugal) and Andalusia (Spain),

Morocco and very rare in Italy (Moreno et al. 2008).

C. Salvador � H. Vicente � J. M. Arteiro �A. T. Caldeira (&)

Department of Chemistry and Evora Chemistry Centre,

University of Evora, Rua Romao Ramalho n859,

7000-671 Evora, Portugal

e-mail: atc@uevora.pt

C. Salvador

e-mail: katia_salvador@hotmail.com

H. Vicente

e-mail: hvicente@uevora.pt

J. M. Arteiro

e-mail: jmsa@uevora.pt

M. R. Martins

Chemistry Department and Institute of Mediterranean

Agricultural and Environmental Sciences, University

of Evora, Rua Romao Ramalho n859, 7000-671 Evora,

Portugal

e-mail: mrm@uevora.pt

J. Neves

Department of Informatics, University of Minho,

Braga, Portugal

e-mail: jneves@di.uminho.pt

123

Agroforest Syst (2013) 87:295–302

DOI 10.1007/s10457-012-9548-y

It is characterised by a large and robust basidium,

with a cap 8–17 cm in diameter, a hemispheric

morphology when young and plane-convex in matu-

rity, and a slight depression in the centre (Fig. 1). The

hymenium is constituted of broad laminas, only

slightly serrated, free or subordinated, with few

lamellae, white but soon going ochraceous and red-

mottled. The stipe is cylindrical, partly smooth to

slightly fibrillose, from 7–13 cm long and 2–5 cm in

diameter, paler in colour than the cap or showing

pinkish-brown hues; it has an unclear annulus, broken

up by the growth of the carpophore, the remainder

like threads surrounding the stipe. The base of the

stem is constituted by a membranous volva, the same

colour as the rest of the fruiting body, able to become,

sac-like, half the height of the stipe. The flesh is firm,

very compact, white, but pinkish when in contact with

the air, with a pungent flavour and odour, like damp

earth. The spores are clumped or scattered and white

when fresh, cream when dry (Moreno-Rojas et al.

2004).

As far as we know there are few studies evaluating

the inorganic composition of A. ponderosa in Iberian

Peninsula. The nutritional values of these fungi make

them highly exportable. Mineral and organic compo-

sition of mushrooms depends on the ecosystem where

they grow. Mushrooms can accumulate high concen-

trations of some elements, namely toxic metals,

because of the symbiotic relation between these

macrofungi and some plants in its habitats (Vetter

2005; Kalac 2010). Therefore it is crucial to determi-

nate their inorganic content in further commercialisa-

tion process.

On the other hand, due to the wide diversity of

mushrooms in nature, it is essential to differentiate and

to identify the various edible species. Genetic profiles

and polymorphic sequences can constitute important

tools for a fast and effective characterisation, namely

in certification processes. In respect to A. ponderosa

mushrooms, existing studies are still incomplete and a

profound knowledge is quite required, especially with

regard to their correct identification and nutritional

value. Data mining tools were used in order to

establish a bridge of inorganic contents, molecular

fingerprints and geographical sites.

The conventional modelling tools have a great

number of drawbacks, since they do not allow the

prediction of singularities inside complex data.

In recent years, some artificial intelligence based

tools, namely Artificial Neural Networks (ANNs) and

Decision Trees (DTs) have been applied for fungal

environment systems (Caldeira et al. 2011a, b).

However, the establishment of a bridge between

inorganic contents, A. ponderosa molecular finger-

prints and geographical sites is a complex and highly

nonlinear problem for which, to our knowledge, no

methods have been reported in the literature.

Artificial Neural Networks (ANNs) are computa-

tional tools which attempt to simulate the architecture

and internal operational features of the human brain

and nervous system. Several types of neural networks

can be found in the literature. However, the multilayer

perceptron is the most common neural network type.

This type of networks are formed by three or more

layers of artificial neurons or nodes, the basic

computing units, which include an input layer, an

output layer and a number of hidden layers with a

certain number of active neurons connected by

feedforward links, to which are associated modifiable

weights. In addition, there is also a bias, which is

connected to neurons in the hidden and output layers.

The number of nodes in the input layer denotes the

number of independent variables and the number of

nodes in the output layer stands for the number of

dependent variables (Haykin 2008).

The aim of the current study was to use artificial

intelligence based tools, particularly ANNs to address

Fig. 1 Macroscopic aspects of edible Mushrooms A. ponder-osa in some maturity stages

296 Agroforest Syst (2013) 87:295–302

123

this problem. ANNs can learn from examples, are fault

tolerant in the sense that they are able to handle noisy

and incomplete data, are able to deal with nonlinear

problems and, once trained, can perform prediction

and generalisation at high speed (Galushkin 2007;

Haykin 2008).

This work is a new approach for the A. ponderosa

species, since until now no study was performed

concerning both trace metals composition and molec-

ular identification.

Materials and methods

Sample collection and preservation

Fruiting bodies of the A. ponderosa mushrooms were

collected in spring, between February and April 2010,

from six different location areas, in the southwest of

the Iberian Peninsula, namely Evora, Beja, Mina de

Sao Domingos, Santo Aleixo da Restauracao and Vila

Nova de Sao Bento (Alentejo, Portugal) and from

Cabezas Rubias (Andalusia, Spain). The mushrooms

under analysis were collected in acid soils, in forests of

Quercus suber, Q. ilex ssp. ballota, Cistus ladanifer

and Cistus laurifolius of the abovementioned sites, at

the same growth stage to avoid the effect of size.

Fruiting bodies were identified by a specialist, based

on morphological features according to taxonomic

description of A. ponderosa (Malencon and Heim

1942).

Three individuals were sampled per location. The

material was weighed and placed in sterile bags for its

inorganic study and molecular characterisation. For

the microsatellite primer M13-PCR molecular study,

the A. ponderosa strains were compared with other

Basidiomycetes (Pleurotus ostreatus and Lactarius

deliciosus) and with one Ascomycete strain (Terfezia

arenaria syn. Tuber arenaria).

Inorganic characterisation

The inorganic contents (Ca, Mg, Fe, Cu, Zn, Mn, Na,

K and P), of A. ponderosa samples were analysed

according to Moreno-Rojas et al. (2004).

Data was evaluated statistically using the SPSS�

16.0 software, by descriptive parameters and by One-

way ANOVA, followed by Tukey test.

Molecular analyses

The genomic DNA extraction from the small frag-

ments of fruiting bodies was performed by the

modified microsphere method (Martins 2004). DNA

amplification was carried out with M13 primer

(50-GAGGGTGGCGGTTCT-30), as described by

Caldeira et al. (2009).

Phylogenetic tree was generated by the UPGMA

method, using the Dice coefficient of similarity.

ANNs

In this study the most common neural network type,

the multilayer perceptron, was adopted. In the training

phase, the backpropagation algorithm (Rumelhart

et al. 1986) was applied. In all experiments, the

sigmoid activation function was used:

u uj

� �¼ 1

1þ e�uj

where uj denotes the weighted sum of the j th neuron

for the input received from a former layer with n

neurons, calculated as:

uj ¼Xn

i¼1

wijxi þ biasj

where wij denotes the weight between the j th neuron

and the i th neuron in the preceding layer, xi denotes

the output of the i th neuron in the preceding layer and

biasj denotes the weight between the j th neuron and

the bias neuron in the preceding layer.

It was used the Waikato Environment for Knowl-

edge Analysis (WEKA) (Hall et al. 2009) to imple-

ment ANNs keeping the default software parameters.

The methodology adopted started with a small

network and continued by adding nodes to improve

performance i.e. in order to minimise an error metric

(Han and Kamber 2006).

To ensure statistical significance of the attained

results, 20 runs were applied in all tests. In each

simulation, the available data was randomly divided into

three mutually exclusive partitions: the training set, with

60 % of the available data, used during the modelling

phase, the test set, with 25 % of the examples, used after

training in order to evaluate the model performance, and

the validation set, with the remaining 15 % of data, to

validate the models (Souza et al. 2002).

Agroforest Syst (2013) 87:295–302 297

123

Results and discussion

Mineral composition

Results of mineral content of the selected A. ponder-

osa strains are described in Table 1.

The ANOVA one-way analysis showed that min-

eral contents of A. ponderosa strains were statistically

significant (p \ 0.001) for all elements investigated.

Tukey post hoc test allows correlating the strains from

different areas and make in evidence homogeneous

groups, based on the similarity of the strains, for each

mineral content (p \ 0.001). Potassium was the

element present in higher concentration in all Amanita

strains, the Cabezas Rubias strains show the highest

value, on the other hand, Mn was the element with

lower values and Santo Aleixo da Restauracao strains

presented the lowest value. Ca contents were signif-

icantly different between A. ponderosa strains from

Santo Aleixo da Restauracao, Beja, Evora and Cabe-

zas Rubias (p \ 0.001), but do not present significant

differences for the strains collected from Mina de Sao

Domingos and Vila Nova de Sao Bento (p = 0.377).

Mg contents were significantly different for mush-

rooms collected from the different areas (p \ 0.001)

except for the strains from Evora and Vila Nova de Sao

Bento (p = 1). Zn, Cu, Fe, Mn and K levels were

significantly different for the six A. ponderosa strains

tested (p \ 0.001). Na content showed significant

differences between location areas, except for Evora

and Mina de Sao Domingos strains (p = 0.607). P

contents also showed significant differences between

location areas, except for Evora and Beja (p = 0.121).

As far as we know there are few studies evaluating

the inorganic composition of A. ponderosa in Iberian

Peninsula. A profound technical knowledge is

required for a product that is so eagerly consumed,

especially with regard to its correct identification and

nutritional value. Trace elements like Fe, Cu, Zn and

Mn play an important role in biological systems (FAO/

WHO 2002); however, they can also produce toxic

effects when the metal intake is excessively (Sesli

et al. 2008). Fe values obtained in this study were

lower than those reported in the literature (Ouzouni

et al. 2007). Cu levels were near those related for

Amanita spp. whereas lower than others edibles mush-

rooms species reported in the literature (Moreno-Rojas

et al. 2004; Sesli et al. 2008). Cu levels were near those

related for Amanita spp. whereas lower than others Ta

ble

1M

iner

alco

nte

nt

of

wil

dA

ma

nit

ap

on

der

osa

mu

shro

om

sco

llec

ted

fro

msi

xd

iffe

ren

tg

eog

rap

hic

alsi

tes

Sam

ple

Am

an

ita

po

nd

ero

sa

San

toA

leix

o

da

Res

tau

raca

o

Vil

aN

ov

a

de

Sao

Ben

to

Bej

aM

ina

de

Sao

Do

min

go

s

Ev

ora

Cab

ezas

Ru

bia

s

Ca

(mg

/Kg

)3

6.7

82

±0

.08

6a

61

.40

5±

0.1

25

b5

5.4

67

±0

.12

6c

61

.20

9±

0.1

23

b5

7.1

99

±0

.11

9d

28

8.0

19

±0

.12

0e

Mg

(mg

/Kg

)4

8.4

08

±0

.02

7a

67

.41

0±

0.0

39

b6

6.2

23

±0

.03

9c

77

.47

5±

0.0

77

d6

7.4

07

±0

.07

4b

10

3.3

97

±0

.07

5e

Zn

(mg

/Kg

)4

.69

3±

0.0

08

a8

.84

0±

0.0

11

b9

.39

5±

0.0

11

c9

.58

5±

0.0

11

d7

.10

4±

0.0

10

e1

0.7

63

±0

.01

1f

Cu

(mg

/Kg

)1

2.9

31

±0

.01

4a

17

.68

6±

0.0

20

b3

3.6

50

±0

.02

0c

35

.11

4±

0.0

20

d1

8.7

27

±0

.01

9e

11

.48

1±

0.0

19

f

Fe

(mg

/Kg

)5

.08

3±

0.0

21

a1

8.6

10

±0

.03

0b

20

.46

9±

0.0

61

c6

.49

1±

0.0

30

d2

7.6

02

±0

.05

8e

15

.79

9±

0.0

29

f

Mn

(mg

/Kg

)4

.72

7±

0.0

06

a7

.33

6±

0.0

08

b7

.57

2±

0.0

08

c1

0.9

28

±0

.00

8d

5.5

94

±0

.00

8e

6.6

21

±0

.00

8f

K(m

g/K

g)

1,8

70

.00

0±

10

.62

5a

63

8.8

67

±3

.07

1b

54

9.7

46

±3

.08

8c

3,5

35

.08

6±

18

.22

2d

3,2

72

.85

5±

17

.59

6e

5,6

87

.65

3±

29

.62

3f

Na

(mg

/Kg

)1

10

.50

0±

1.0

63

a1

50

.50

2±

1.5

36

b1

38

.98

1±

1.5

44

c1

15

.40

7±

1.5

19

a1

17

.30

7±

1.4

66

a2

16

.24

9±

1.4

81

d

P(m

g/K

g)

3.3

79

±0

.07

4a

20

.17

1±

0.1

38

b1

9.1

70

±0

.03

9c

15

.67

6±

0.1

00

d1

8.9

64

±0

.09

7c

5.9

80

±0

.04

9e

No

teA

lld

ata

are

sho

wn

asav

erag

em

edia

nv

alu

es(n

=1

8)

±S

E

Dif

fere

nt

sup

ersc

rip

tle

tter

sin

the

sam

ela

ne

ind

icat

esi

gn

ifica

nt

dif

fere

nce

sat

99

.9%

(p\

0.0

01

)le

vel

(AN

OV

A,

Tu

key

test

)

298 Agroforest Syst (2013) 87:295–302

123

edibles mushrooms species reported in the literature

(Sesli et al. 2008; Moreno-Rojas et al. 2004). Present

Cu concentrations in mushrooms are not considered a

health risk (Commission 2003). Zn is an element with

biological properties that can be accumulated by

mushrooms (Mendil et al. 2004). Results showed that

Zn and Mn content were near for those related by

Moreno-Rojas et al. (2004) for A. ponderosa strains

although lower than those reported for others edible

mushrooms (Kalac and Svoboda 2000).

Molecular analyses

M13-PCR fingerprinting of different A. ponderosa

strains assayed generated different patterns with 8–20

bands, ranging from 260 to 1,490 bp. A. ponderosa

band profiles were compared with other edible mush-

rooms namely Pleurotus ostreatus and Lactarius

deliciosus and with one wild edible truffle Terfezia

arenaria. Dendrogram molecular analysis shows three

different clusters (Fig. 2). Cluster A illustrate that T.

arenaria, an ectomycorrhizal ascomycete, is the most

distant species from the remaining strains (40 % of

similarity). Cluster B was obtained for L. deliciosus

and P. ostreatus (similarity of 49 %) and cluster C

grouped all A. ponderosa strains (67 % of similarity).

This approach although producing band profiles very

similar for the A. ponderosa strains is able to detect

intraspecific variation, showing different polymor-

phisms between strains of the same species. The most

similar strains, from Evora and Santo Aleixo da

Restauracao, forming a cluster with 85 % of similar-

ity, while strains collected from Beja, Vila Nova de

Sao Bento and Cabezas Rubias give another cluster

with 80 % of similarity (Fig. 2).

Literature related a range of molecular approaches

based on RAPD analyses in order to characterise and

identify edible mushrooms (Moncalvo et al. 2000;

Firenzuoli et al. 2007), however, a lack of information

to A. ponderosa mushrooms is verified, only one study

is reported to distinguish A. curtipes and A. ponderosa

species sequencing the D1–D2 domains of the 28S

rRNA gene and ITS1-5.8S-ITS2 region (Moreno et al.

2008). Previous studies performed with seven micro-

satellite primers showed that the M13-PCR approach

is a rapid and precise method that allows differenti-

ation at the species and strain level (Caldeira et al.

2009).The M13 primer were also used to study genetic

variability in yeasts and filamentous fungi species

(Alves et al. 2007; Lopes et al. 2007). Actually, M13-

PCR technique is easy to implement, less time-

consuming than other molecular approaches with

restriction enzymes. It is highly reproducible and is

much more robust than RAPD-PCR for species

Fig. 2 Dendrogram analysis based on the M13-PCR for A.ponderosa strains and other wild edible mushrooms

Table 2 The main fields in the dataset

Mineral

composition

DNA fragments

(bp)

Location area

Ca 422.54 Evora

Mg 609.77 Beja

Zn 723.42 Mina de Sao

Domingos

Cu 786.26 Santo Aleixo da

Restauracao

Fe 856.72 Vila Nova

de Sao Bento

Mn 1,090.20 Cabezas Rubias

K 1,189.66

Na

P

Agroforest Syst (2013) 87:295–302 299

123

identification by visual analysis of the amplification

patterns.

Modelling

Database

The data used in this study containing a total of 108

records with 17 fields. Table 2 shows a synopsis of the

relevant fields of the dataset. The main fields in the

dataset were the mineral composition of A. ponderosa

strains, M13-PCR band profile and the locations in

which the samples were collected. For the molecular

analysis characterisation, the indicators chosen were

the presence or the absence of the bp DNA fragments

(Table 2).

ANN model

Artificial neural networks were used to predict the

M13-PCR DNA band profile from A. ponderosa based

on inorganic composition of the mushrooms.

The ANN model obtained is showed in Fig. 3. The

architecture of the model consists in an input layer

with six nodes, a hidden layer with seven and a 14

nodes output layer.

It should be emphasised that the algorithm chooses

only six mineral composition variables (Cu, Fe, K,

Mn, Na, Zn), even though all the mineral composition

variables presented in Table 1 were available for use.

In order to evaluate the model output sensitivity to

changes in its input variables it was used the sensitiv-

ity, according to the variance (Kewley et al. 2000), to

compute the relative importance of the input variables.

The results are presented in Fig. 4 and reveal that the

most informative variable is Zn followed by Fe and

Cu. These results seem to suggest that these three

variables have direct relevance and play a significant

role in the M13-PCR DNA band profile from A.

ponderosa. Table 3 presents the coincidence matrix

for the model. The values denote the average of the 20

Fig. 3 Artificial neural networks model Fig. 4 Relative importance of input variables

300 Agroforest Syst (2013) 87:295–302

123

runs. The results reveal that the model exhibits

accuracy of 100 % in predicting M13-PCR DNA

band profile from A. ponderosa for training set, test set

and validation set.

Conclusions

M13-PCR allows to distinguish the molecular profiles

from A. ponderosa strains collected from different

location areas and the inorganic analyses showed that

mineral composition of mushrooms depends on the

ecosystem where they grow. The results show that it was

possible to correlate the molecular and inorganic data.

It would be interesting to study, in parallel, the

concentration of minerals in soils from the different

locations in order to distinguish substrate effects from

inter-individual or inter-strain differences.

The present findings are wide potential application

and both health and economical benefits arise from

this study.

References

Alves A, Phillips A, Henriques I, Correia A (2007) Rapid dif-

ferentiation of species of Botryosphaeriaceae by PCR fin-

gerprinting. Res Microbiol 158:112–121

Caldeira AT, Salvador C, Pinto F, Arteiro JM, Martins MR

(2009) MSP-PCR and RAPD molecular biomarkers to

characterize Amanita ponderosa mushrooms. Ann Micro-

biol 59(3):1–6

Caldeira AT, Vicente H, Arteiro JM, Roseiro JC, Neves J

(2011a) An artificial intelligence approach to Bacillusamyloliquefaciens CCMI 1051 cultures: application to the

production of antifungal compounds. Bioresour Technol

102:1496–1502

Caldeira AT, Roseiro JC, Arteiro JM, Neves J, Vicente H

(2011b) Production of bioactive compounds against wood

contaminant fungi: an artificial intelligence approach. In:

Caldeira Fernando (ed) Minimizing the environmental

impact of the forest products industries. University Fer-

nando Pessoa, Porto, pp 131–137

Commission E (2003) Opinion of the scientific committee on

food on the tolerable upper intake level of copper. Health

and Consumer Protection Directorate-General, Brussels

FAO/WHO (2002) Human vitamin and mineral requirements.

World Health Organization, Food and Agriculture Orga-

nization of United Nations, Rome

Firenzuoli F, Gori L, Lombardo G (2007) The medicinal

mushroom Agaricus blazei Murrill: review of literature and

pharmaco-toxicological problems. Adv Access Pub 5(1):

3–15

Galushkin AI (2007) Neural networks theory. Springer, New

York

Hall M, Frank E, Holmes G, Pfahringer B, Reutemann P, Witten

IH (2009) The WEKA data mining software: an update.

SIGKDD Explor 11:10–18

Han J, Kamber M (2006) Data mining: concepts and techniques.

Morgan Kauffmann Publishers, San Francisco

Haykin S (2008) Neural networks and learning machines. Pre-

ntice Hall, New Jersey

Kalac P (2010) Trace element contents in European species of

wild growing edible mushrooms: a review for the period

2000–2009. Food Chem 122:2–15

Kalac P, Svoboda L (2000) A review of trace element concen-

trations in edible mushrooms. Food Chem 69:73–281

Kewley R, Embrechts M, Breneman C (2000) Data strip mining

for the virtual design of pharmaceuticals with neural net-

works. IEEE T Neural Networ 11:668–679

Lopes M, Silva D, Freitas G, Tenreiro R (2007) Simultaneous

identification and typing of Candida species by MSP-PCR

and AFLP: study of clinical isolates from a Portuguese

pediatric hospital. J Mycol Med 17:157–167

Malencon G, Heim R (1942) Notes critiques sur quelques

hymenomycetes d’Europe et d’Afrique du Nord I. Les

amanites blanches meridionales. Bull Soc Mycol Fr 58:

14–34

Martins MR (2004) Degradacao Biologica de Fungicidas em

Amostras de Solo, Tese de Doutoramento. Universidade de

Evora, Evora

Mendil D, Uluozlu O, Hasdemir E, Caglar A (2004) Determi-

nation of trace elements on some wild edible mushroom

Table 3 The coincidence matrix for ANN model

Training set Test set Validation

A P A P A P

1,190 bp DNA fragment

A 30 0 16 0 8 0

P 0 34 0 12 0 8

1,090 bp DNA fragment

A 53 0 24 0 13 0

P 0 11 0 4 0 3

857 bp DNA fragment

A 21 0 11 0 4 0

P 0 43 0 17 0 12

786 bp DNA fragment

A 53 0 24 0 13 0

P 0 11 0 4 0 3

723 bp DNA fragment

A 53 0 24 0 13 0

P 0 11 0 4 0 3

610 bp DNA fragment

A 21 0 11 0 4 0

P 0 43 0 17 0 12

423 bp DNA fragment

A 41 0 20 0 11 0

P 0 23 0 8 0 5

Agroforest Syst (2013) 87:295–302 301

123

samples from Kastamonou, Turkey. Food Chem 88:

281–285

Moncalvo J, Drehmel D, Vilgalys R (2000) Variation in modes

and rates of evolution in nuclear and mitochondrial ribo-

somal DNA in the mushroom genus Amanita (Agaricales,basidiomycota): phylogenetic implications. Mol Phyloge-

net Evol 16:8–63

Moreno G, Platas G, Pelaez F, Bernedo M, Vargas A, Daza A,

Santamarıa C, Camacho M, de la Osa L, Manjon J (2008)

Molecular phylogenetic analysis shows that Amanitaponderosa and A. curtipes are distinct species. Mycol

Progress 7:41–47

Moreno-Rojas R, Dıaz-Valverde A, Moreno-Arroyo B, Gon-

zalez T, Capote C (2004) Mineral content of gurumelo

(Amanita ponderosa). Food Chem 85:325–330

Ouzouni P, Veltsistas P, Paleologos E, Riganakos K (2007)

Determination of metal content in wild edible mushroom

species from regions of Greece. J Food Compos Anal

20:480–486

Rumelhart D, Hinton G, Williams R (1986) Learning internal

representation by error propagation. In: Rumelhart DE,

McCleland JL (eds) Parallel distributed processing. MIT

Press, Massachusetts, pp 318–362

Sesli E, Tuzen M, Soylak M (2008) Evaluation of trace metal

contents of some wild edible mushrooms from Black sea

region, Turkey. J Hazard Mater 160:462–467

Souza J, Matwin S, Japkowicz N (2002) Evaluating data mining

models: a pattern language. In: Proceedings of the 9th con-

ference on pattern language of programs, University of Illi-

nois at Urbana-Champaign, Urbana, Illinois, USA, 8-12

Sept 2002, pp 1–23

Vetter J (2005) Mineral composition of basidiomes of Amanita

species. Mycol Res 109(6):746–750

302 Agroforest Syst (2013) 87:295–302

123