Volatile fingerprinting of ripened cheese for authentication and characterisation of different dairy systems

Figures & data

Table 1. Descriptive statistics (SD between brackets) of productive traits, and milk and cheese composition of sampled cows according to different dairy systems.

	5 DAIRY SYSTEMS^a:					3 DAIRY SYSTEMS^a:
	Traditional		Modern				Modern
				TMR			No-silage^c	Silage^d
	No-AF	AF	No-TMR	No-silage	Silage	Traditional^b
Herds, n	9	12	30	14	7	21	44	7
Cow, n	135	180	446	209	105	315	655	105
Animal condition	tied	tied	loose	loose	loose	tied	loose	loose
Milking	at stalls	at stalls	parlor	parlor	parlor	at stalls	parlor	parlor
Major forage	hay	hay	hay	hay	hay/silage	hay	hay	hay/silage
Major concentrate	compound	compound	compound	cereal mix	cereal mix	compound	compound	cereal mix
Forage:concentrate	0.73:0.27	0.65:0.35	0.64:0.36	0.52:0.48	0.47:0.53	0.69:0.31	0.61:0.39	0.47:0.53
Productive traits
Milk yield, kg × d^–1	19.5(±7.2)	21.8(±6.4)	24.7(±7.0)	28.9(±8.3)	26.9(±7.6)	20.8(±6.92)	26.1(±7.73)	26.9(±7.6)
Days in milk, d	187(±145)	201(±135)	203(±138)	195(±141)	195(±126)	195(±140)	201(±139)	195(±126)
Parity, n	3.05(±2.11)	2.94(±2.05)	2.76(±1.75)	2.47(±1.56)	2.42(±1.51)	2.99(±2.07)	2.66(±1.67)	2.42(±1.51)
Milk composition
Protein, %	3.66(±0.45)	3.67(±0.43)	3.77(±0.41)	3.81(±0.43)	3.87(±0.39)	3.67(±0.44)	3.78(±0.41)	3.87(±0.39)
Fat, %	4.26(±0.84)	4.24(±0.74)	4.37(±0.95)	4.36(±0.79)	5.03(±1.15)	4.25(±0.78)	4.37(±0.90)	5.03(±1.15)
SCS^5, U	3.14(±2.07)	2.91(±2.05)	3.15(±1.75)	2.84(±1.79)	2.99(±1.82)	3.01(±2.06)	3.05(±1.77)	2.99(±1.82)
pH	6.60(±0.12)	6.66(±0.08)	6.63(±0.07)	6.64(±0.07)	6.63(±0.10)	6.63(±0.10)	6.64(±0.07)	6.63(±0.10)
Cheese composition
Protein, %	26.8(±4.3)	27.1(±4.1)	27.5(±4.1)	27.5(±4.0)	26.0(±4.2)	26.9(±4.2)	27.5(±4.1)	26.0(±4.2)
Fat, %	38.8(±4.6)	38.2(±4.1)	37.7(±4.3)	37.9(±4.5)	39.8(±4.9)	38.5(±4.3)	37.8(±4.3)	39.8(±4.9)
pH	5.14(±0.19)	5.17(±0.18)	5.13(±0.20)	5.17(±0.13)	5.21(±0.14)	5.16(±0.18)	5.15(±0.18)	5.21(±0.14)

^aAF: automatic feeders at mangers to control individually concentrate distribution; TMR: total mixed ration; modern TMR no silage = water added in the mixer wagon to enhance mixing; productive traits as well as milk and cheese composition are expressed as mean and standard deviation (in parenthesis).

^b Traditional: cluster of herds composed by traditional dairy systems with and without AF.

^cModern no silage: cluster of herds composed by modern dairy systems with hay plus compound feed and modern TMR without silage.

^dModern silage: cluster of herds that used TMR and corn silage; ⁵SCS = log₂(SCC/100,000) + 3.

Table 2. Correct classification (% average of 10-fold cross-validation) of the linear discriminant analysis (LDA) based on volatile organic compounds of cheeses measured by PTR-ToF-MS to authenticate the dairy systems.

LDA^b		5 DAIRY SYSTEMS^a						3 DAIRY SYSTEMS^a
		Traditional		Modern			TOTAL		Modern
Training-set	Testing-set	No-AF	AF	No-TMR	TMR
					No-silage	Silage		Traditional^c	No-silage^d	Silage^e	TOTAL
Cows LDA
Cow-All	Cow-All	65	66	82	63	77	73	69	90	68	82
Cow-Cal	Cow-Cal	70	71	82	67	81	75	71	90	73	83
Cow-Cal	Cow-Val	40	49	58	38	37	48	57	77	31	66
Herds LDA
Herd-All	Herd-All	100	83	93	86	100	92	91	93	100	93
Herd-Cal	Herd-Cal	92	90	91	83	100	92	88	95	98	94
Herd-Cal	Herd-Val	13	11	49	23	10	28	25	70	10	51
Crossed LDA
Cow-Cal	Herd-Val^f	100	67	100	100	100	95	67	100	100	90
Herd-Cal	Cow-Val^g	35	38	35	31	48	36	42	59	52	53

^aAF: automatic feeders at mangers to control individually concentrate distribution; TMR: total mixed ration; modern TMR no silage = water added in the mixer wagon to enhance mixing.

^bLinear discriminant analysis performed without validation (All) and with validation where ‘Cal’ and ‘Val’ are the calibration and validation, respectively.

^cTraditional: cluster of herds composed by traditional dairy systems without and with AF.

^dModern no silage: cluster of herds composed by modern dairy systems with hay plus compound feed and modern TMR no silage.

^eModern silage: cluster of herds that used TMR and corn silage.

^fThe linear discriminant functions calculated with the cow training data set (Cow-Cal) have been validated using the herd testing data set (Herd-Val).

^gThe linear discriminant functions calculated with the herd training data set (Herd-Cal) have been validated using the cow testing data set (Cow-Val).

Table 3. Variance explained by each linear discriminant function (LD) of the 12 analyses (LDA) according 3 or 5 dairy systems.

LDA^a		5 Dairy Systems				3 Dairy systems
Training-set	Testing-set	LD1	LD2	LD3	LD4	LD1	LD2
Cows LDA
Cow-All	Cow-All	38.3	24.5	22.6	14.6	58.4	41.6
Cow-Cal	Cow-Cal	37.6	25.2	22.4	14.8	58.1	41.9
Cow-Cal	Cow-Val	35.8	32.9	17.2	14.1	58.3	41.7
Herds LDA
Herd-All	Herd-All	40.9	24.5	18.3	16.4	63.2	36.8
Herd-Cal	Herd-Cal	37.9	25.1	22.4	14.6	60.2	39.8
Herd-Cal	Herd-Val	78.5	10.0	8.9	2.6	78.6	21.4

^aThe name of the model refers to the source of information (individual cow or herd average volatile organic compounds fingerprint) and to the dataset used for calibration (the first) and for validation (the second); All: the whole dataset; Cal: the dataset used for calibration (about 75% of All); Val: the dataset used for validation (about 25% of All).

Figure 1. Bi-plot score of the linear discriminant functions (LD1; LD2) obtained using the full data set (All) including the 1,075 cows (a) and 72 herds (b) as observation and the spectrometric peaks that characterised the volatile profile of individual model cheeses as variables. The cows and the herd are classified in three dairy systems.

Figure 2. Bi-plot score of the linear discriminant functions (LD1; LD2) obtained using the full data set (All) including the 1,075 cows (a) and 72 herds (b) as observations and the spectrometric peaks that characterised the volatile profile of individual model cheeses as variables. The cows and the herds are classified in five dairy systems. Traditional: cluster of herds composed by traditional dairy systems without and with AF (Automatic Feeders at mangers to control individually concentrate distribution); Modern no silage: cluster of herds composed by modern dairy systems with hay plus compounds feed and modern TMR (total mixed ration) no silage; Modern silage: cluster of herds that used TMR and corn silage; CF: compounds feed.

Table 4. Coefficients in the linear discriminant function (LD) of the 12 analyses (LDA) and average content for the most discriminant spectrometric peaks (having coefficients >0.35) of the 1,075 model cheeses (cows) belonging to 3 and 5 dairy systems^a.

Measured mass, m/z	Sum formula	Theoretical mass, m/z	Tentative VOC identification^b	Average content ppbv^3	3 Dairy systems^a					5 Dairy systems^a
					LDA:		Traditional^d	Modern		LDA:		Traditional		Modern
					LD1	LD2		No-silage^e	Silage^f	LD1	LD2			No-TMR	TMR
												No-AF	AF		No-silage	Silage
61.028	C2H4O2^+	61.0284	Acetic acid, acetate ester fragment	11.04	–1.01		++	––	+	1.00		–	++	+	––	+
71.086	C5H11^+	71.0855	3-Methyl-butan-1-ol, 3-methyl-3-buten-1-ol, pentan-1-ol	10.38	–0.83		––	+	++	0.85		–	––	+	+	++
77.060	C3H9O2^+	77.0597	1,2-Propanediol	6.43			–	++	––		–0.64	–	+	++	–	––
83.071			NA	7.44	0.93	–0.81	––	+	++	–0.55	–1.30	–	––	–	++	+
85.065	C5H9O^+	85.0653	Methylbutanoic acid fragment, pentenal	6.60		–0.60	––	+	++			––	++	+	–	+
87.080	C5H11O^+	87.0804	3-Methylbutanal, 2-methylbutanal, pentan-2-one	10.42			––	++	–		0.89	–	––	+	++	–
93.090			NA	6.23	–0.45		––	+	++	–0.50	–0.68	–	–	––	++	+
97.101	C7H13^+	97.1012	Heptanal fragment	6.23		0.97	––	++	–		1.18	––	–	+	++	–
101.060	C5H9O2^+	101.0597	2,3-Pentanedione	6.36			––	+	++	0.42		–	––	+	–	++
101.097	C6H13O^+	101.0970	Hexan-2-one, hexanal	6.38	–1.06	–1.05	––	+	++	1.29	–0.89	––	–	–	++	+
103.075	C5H11O2^+	103.0754	3-Methylbutanoic acid (isovaleric acid)	7.76		0.70	––	+	++			––	++	+	–	+
109.070	C2^[ac]CH10O3N^+	109.0760	2,6-Dimethylpyrazine	6.30			+	––	++		–0.64	–	++	+	––	+
115.112	C7H15O^+	115.1120	Heptan-2-one, heptanal	8.78	0.69		––	++	–	–0.62	–0.60	––	–	+	++	–
119.107			NA	6.36		–0.82	+	––	++		–0.76	+	–	––	++	+
123.117	C8H15O^+	127.1123	Nonenal fragment	4.82	0.86		––	+	++	–0.59		––	+	+	–	++
129.127	C8H17O^+	129.1274	Octan-1-one, octanal	4.95	–0.47	–0.80	––	+	++	0.37	–0.62	––	–	+	+	++
137.132	C10H17^+	137.1332	Terpenes (p-cymene)	5.28	0.61		++	–	––	–0.62		++	+	+	––	–
141.129	C9H17O^+	141.1270	Unsaturated aldehydes and ketones	5.52		–0.60	––	+	++			––	–	+	+	++
143.143	C9H19O^+	143.1430	Nonan-2-one, nonanal	7.11		0.70	––	+	++			––	–	+	+	++
145.123	C8H17O2^+	145.1223	Octanoic acid	7.43		0.63	–	+	––			––	+	–	++	–

^aAF: automatic feeders at mangers to control individually concentrate distribution; TMR: total mixed ration; without silage: water added in the mixer wagon to enhance mixing; ++ and –– indicate the maximum and minimum volatile compounds concentration; + and – indicate the volatile compounds concentration above or below the average value.

^bNA: not attributed.

^cSpectrometric peaks concentration expressed as natural logarithm of the parts per billion by volume (ppb_v) from 1,075 individual model cheeses.

^dTraditional: cluster of herds composed by traditional dairy systems without and with AF.

^eModern no silage: cluster of herds composed by modern dairy systems with hay plus compounds feed and modern TMR no silage.

^fModern silage: cluster of herds that used TMR and corn silage.

Bargo F, Muller LD, Delahoy JE, Cassidy TW. 2002. Performance of high producing dairy cows with three different feeding systems combining pasture and total mixed rations. J Dairy Sci. 85(11):2948–2963.

 PubMed Web of Science ®Google Scholar