The quality parameters (linearity range, LOD and recovery %) of the major volatile compounds quantified from the pineapple breads are shown in Table 1. The recoveries were good (>87.8%). This was followed by an excellent linearity in all cases. The limits of detection (LOD) which is the lowest concentration at which the results still satisfy some predetermined acceptance criteria or the mean blank value plus (3 × the relative standard deviation of the analytical blank values) [17] were low enough to determine the aroma compounds in the pineapple breads.

The sensory evaluation showed appreciable differences among the different pineapple bread types (Table 2). The best results were obtained in the case of the partially baked pineapple bread (Ppb), while the conventionally baked pineapple bread (Cpb) recorded the lowest scores. Statistical analysis (one-way ANOVA) showed that the scores for aroma and taste between all types of pineapple breads were not significantly different (p > 0.05). On the contrary, the scores for the overall quality between partially baked pineapple bread and conventionally baked bread showed significant (p < 0.05) differences. The overall quality is that sensational attributes that makes a product fundamentally different from others. Some of the factors contributing towards quality of food are; appearance, colour, taste, aroma, adulterants and nutritional value. While the Ppb produced higher intensity of dough and sourdough flavour with slight fruity notes, the Cpb and Fpb had lower flavour nuances.

To elucidate the molecular principles responsible for the observed aroma impressions, the flavour volatile compounds of the pineapple breads were first isolated by means of headspace solid-phase microextraction and were subsequently analysed by means of gas chromatography-olfactometry, as well as mass spectrometry. This method and the application of the aroma extract dilution analysis (AEDA) led to the detection of 59 odour-active compounds in the FD factor range of 2–128. These compounds, together with their odour impressions and retention indices are shown in Table 3.

The complex chemical and enzymatic reactions which occurred during fermentation and baking gave rise to a number of volatile and non-volatile compounds. It is likely that these compounds in no doubt contributed to the overall flavour of the final pineapple bread. In this study, the identified volatile compounds comprised eight acids, eight alcohols, 12 aldehydes, 12 ketones, 10 esters and three furan derivatives (Table 3). In general, all main classes of compounds commonly listed as thermally generated flavours in baked breads were identified in the pineapple breads (Cpb, Fpb and Ppb). The flavour compounds revealed fruity (nos. 1, 5, 8, 15, 17, 26, 46, 56, and 58), buttery (nos. 4, 7 and 45), roasty (nos. 29, 35 and 37), sweaty (nos. 14, 23, 28 and 50), fatty (nos. 27, 33, 42, 43, 44, 48 and 54) odour qualities. Also, diverse sweet, malty, vanilla- or honey-like/flowery compounds were detected (nos. 9, 18, 25, 30, 36, 38, 49, 57 and 59) (Table 3). It is noteworthy that aroma compound number 20 (Table 3) with popcorn-like aroma failed to yield any signal at the flame ionization detector. This indicated a very low odour threshold for this compound. This compound was later characterized by co-injection with the reference compound on the DB-5 column (Table 3) and its odour quality at the sniffing port revealed it to be 2-acetyl-1-pyrroline.

The production of volatile compounds is generally influenced by dough fermentation [24], proofing and baking [25]. For instance, prolonged dough fermentation has been reported to increase the concentration of some volatile compounds such as 3-methylbutanol, 2-phenylethanol and ethanol [26]. These compounds are directly linked to the fermentative activity of the yeast in the dough fermentation step [27]. In the present study, compounds with appreciably high concentrations were ethanol, ethyl acetate, 2,3-butanedione, 3-hydroxy-2-butanone and ethyl propionate (Table 4). Interestingly, the partially baked pineapple bread (Ppb) recorded the highest concentration of odourants in all cases (Table 4). This is probably due to the slightly higher fermentation time of the Ppb, which might have produced appreciable amounts of free amino acids. The free amino acids can act as precursors of the Strecker reaction leading to the increases obtained in odourant’s concentration [26].

In order to examine the similarities and differences between the three pineapple breads (Cpb, Fpb and Ppb) in terms of the volatile compounds associated with each sample, principal component analysis (PCA) was employed. PCA provided easy visualisation of the complete data set in a reduced dimension plot, showing variability between volatile- and non-volatile compounds of breads. This method was employed to establish relationships between pineapple breads (Cpb, Fpb and Ppb) and their flavour volatile compounds (Table 4). Based on the bread samples’ grouping from PCA, a partial least square discriminant analysis (PLS-DA) was established (Figure 1). The scatter plot of scores of the first two components (in the PLS-DA which explained 92.53% of the total variance in the data) showed the distinctions among the different pineapple breads.

The PLS-DA model (which was based on three replicates and three breads) was however, validated using the response of permutation test through 100 permutations (Figure 2). A model is said to be valid when the intercept of R2 is <0.3 and intercept of Q2 is <0.05 [25]. In this study, the R2 is 0.271 and Q2 is −0.182 respectively. This plot (Figure 2) strongly indicates that the original model is valid because the regression line of the Q2-point intersects the vertical axis on the left and below zero.

Pineapple breads (Cpb, Fpb and Ppb) were separated according to their processing methods (Figure 1). While conventionally baked pineapple bread (Cpb) was situated in the area of positive components 1 and 2, the fully frozen baked pineapple bread (Fpb) was within the area of positive component 1 and negative component 2 respectively. However, partially baked pineapple bread (Ppb) was situated in the area of negative component 1 and positive component 2 respectively. The inter-relationship between the pineapple breads (Cpb, Fpb and Ppb) and the volatile- and non-volatile compounds were examined by the PLS-regression coefficients of scaled and centred variables (Figure 3A–C). Meanwhile, the Cpb showed significant (p < 0.05) positive correlation with hexanal and 4-hydroxy-2,5-dimethyl-3(2H)-furanone, it however, revealed strong negative correlations with benzoic acid, benzaldehyde, and ethyl propionate (Figure 3A). On the other hand, Fpb, revealed strong positive correlations with lactic acid, benzoic acid, benzaldehyde and ethyl propionate (Figure 3B). Moreover, the Ppb showed positive correlations with most of the volatile- and non-volatile compounds respectively (Figure 3C). It however, showed negative correlations with only two compounds namely; benzaldehyde and ethyl propionate. As can be seen, while the alcohols (ethanol, 2/3-methyl-1-butanol and 2-phenylethanol) and the acids showed high positive correlations in the partially baked pineapple bread (Ppb), they however, revealed negative correlations in both conventional and fully frozen baked pineapple breads (Figure 3A–C). Previous studies on sourdoughs [26,27,28] have shown that the presences of acids in sourdoughs are positively valued, as they strengthen the flavour.

As the conventional pineapple bread (Cpb) and the fully frozen pineapple bread (Fpb) were based on the same recipe and fermentation regime, the differences identified in their volatile compounds may be linked to the freezing and thawing stages that distinguished them.

All experiments were carried out with flour purchased from a local market in Malaysia. Its composition was: 12% protein, 13.5% moisture basis and 77% starch (determined by a Megazyme starch kit, Megazyme Int., Wicklow, Ireland), of which 5% was damaged starch (determined by a Megazyme damaged starch kit). Fresh pineapple fruits (Ananas comosus L) of Josephine cultivar with similar characteristics of ripening (skin colour, flat eyes and Brix) were used.

The following odourants: 97% 2-methyl-1-butanol, 96% 3-methyl-1-butanol, 98% 2-phenyl ethanol, 98% 1-butanol, 96% acetic acid, 98% propanoic acid, 97% hexanoic acid, 96% 4-hydroxy-butanoic acid, 89% benzoic acid, 97% hexanal, 98% benzaldehyde, 98% (E,E)-2,4-decadienal, 97% furfural, 98% furfuryl alcohol, 98% ethyl acetate, 98% ethyl propanoate, 97% methyl ethanoate, 97% ethyl octanoate, 98% 2-methylpropanoic acid, 96%; were obtained from (Aldrich, Steinheim, Germany); 99% 2,3-butanedione, 98% 4-hydroxy-2,5-dimethyl-3(2H)-furanone, 98% 3-hydroxy-2-butanone were purchased from Fluka (Neu-Ulm, Germany). Stock standard solutions of 10³ or 10⁴ mg L⁻¹ of each component were prepared by dissolving the pure standard in 40% (v/v) ethanol. The samples were stored at 4 °C. Working standard solutions were prepared daily by mixing an aliquot of each individual solution and diluting with ultra pure water (Millipore Co., Bedford, MA, USA) to obtain a desired concentration.

The recipes and the production processes of the pineapple breads are as shown in Figure 4 [15]. For the preparation of the pineapple breads, three different methods (conventional, fully baked and frozen and partially baked) were employed. The dough was mixed in a KN-200 mixer (Taisho, Denki Co., Ltd., Shiga, Japan) for 7 min at 380 rpm. The dough was flattened with a roller to obtain a round, symmetrical and homogenous thick sheet. It was then divided into 20 samples of 150 g pieces and allowed to rest for 20 min at room temperature. The dough pieces were divided into two batches (i.e., 10 samples of 150 g pieces).

One batch was fermented for 2 h at 35 °C and 85 relative humidity (RH) while the second batch was fermented for 2.15 h at 35 °C and 85 RH. After fermentation, the first batch was baked at 230 °C for 20 min with an initial 0.5 L steam at the start of baking in a ventilated oven. Baked samples were further divided into two groups. A group is referred to as the conventional pineapple bread (Cpb) while the other group was frozen at −30 °C for 30 min and latter defrosted at room temperature for 60 min. This group is referred to as the fully baked frozen pineapple bread (Fpb).

The second fermented batch was partially baked at 190 °C for 3 min with 0.5 L steam at the start of baking similarly to the first batch. The baked bread sample was cooled for 30 min at room temperature and final baking was carried out in a static oven at 230 °C for 12 min. This sample is referred to as partially baked pineapple bread (Ppb).

The sensory acceptability of the pineapple breads (Cpb, Fpb and Ppb) were determined using 24 University Putra Malaysia students whose ages ranged from 19 to 30 years. The students were made up of 12 females and 12 males, respectively. The samples were coded with a three-digit randomised number. The panellists were provided with water as mouth rinse. Each partitioned booth in the taste panel section was lighted with natural white fluorescent light. Panellist scored the breads according to a preference protocol including a scale from 0 (unacceptable) to 10 (excellent) [29]. Odour intensity was monitored using an odour intensity scale, which provides a verbal description of an odour sensation to which a numerical value is assigned (i.e., 0 = no odour; 1 = very weak; 2 = weak; 3 = distinct; 4 = strong; 5 = very strong and 6 = intolerable) [7].

Total titratable acidity (TTA) was measured using the method described by Martinez-Anaya and others [30] and lactic acid was determined using the method described by Quilez and other [31].

A cross sectional cut (10 g of each bread samples) obtained from the central part was mixed together with NaCl (10 mL, 1 g·mL⁻¹) and shaken for 30 s. Five grams of this mixture was placed in a 125 mL flask with a magnetic stirring bar and 5 µL of internal standard and closed with cap. The flask was incubated for 60 min in a heated tray oven at 40 °C. An SPME fibre with two different coatings (75 µm carboxen^TM /polydimethylsiloxane; CAR/PDMS stablFlex^TM) (Supelco Co., Bellefonte, PA, USA) was inserted in the headspace of the vial for 10 min to collect the volatiles. The type of fibre used can often affect the selectivity of the extraction and by using a mixed type; good selectivity is obtained for non-polar analyte as well as for polar analyte. This type of fibre was used successfully in the past for the characterisation of the aroma compounds of partially baked bread [11]. After each extraction, the fibre was inserted into the GC injector port using a 0.75 mm i.d liner (in order to improve the GC resolution). Desorption time and temperature were 5 min and 250 °C respectively. All experiments and sample measurements were carried out in triplicate and the average values were recorded.

The identification and quantification of volatile compounds was carried out with a Shimadzu (Kyoto, Japan) QP-5050A GC-MS instrument equipped with a GC-17A Ver. 3 gas chromatograph with a flame ionization detector (FID). The column was a DB-5 column (30 m × 0.32 mm i.d., film thickness 0.25 µm J & W Scientific, Folsom, CA, USA). Helium was used as carrier gas at a flow rate of 1.5 mL min⁻¹, injection temperature, 250 °C; detector temperature, 280 °C; temperature program commenced at 50 °C and held for 3 min, then raised to 250 °C at a rate of 15 °C min⁻¹, held for 10 min and then increased to 280 °C at a rate of 10 °C min⁻¹, with a final hold time of 5 min. The effluent from the capillary column was split into 2:1 (by vol.) onto two uncoated but deactivated fused silica capillaries (50 cm × 0.32 mm) leading to a FID and a sniffing port.

The mass spectrometer was operated in electron impact mode with the following conditions. The source temperature was 250 °C; the quadruple temperature selected was 280 °C and the relative electron multiplier voltage (EM) applied was 400 V with a resulting voltage of 1,553 V. In order to improve the detection limits, the selected ion monitoring (SIM) mode was used. Compounds identification was based on comparison of linear retention indices (RI), mass spectra (comparison with standard MS spectra databases: Wiley 6), and injection of standards. The quantification was performed using 1-butanol as the added standard. The concentrations of volatile compounds were expressed in microgram equivalents of 1-butanol per gram of bread.

Each bread sample (dry weight) (400 g) was mixed with water (1,800 mL) in a 4-litre distilling flask. The distilling flask was heated in a water bath at 40 °C and the volatile compounds were steam distilled under vacuum (30 mbar) for 5 h. The distillate was extracted with dichloromethane, dried with sodium sulphate and concentrated to 50 µL [26].

GC-olfactometry analysis was performed on a Thermo Scientific GC instrument (Thermo Fisher Scientific, Rivoltana, Italy) equipped with a split-splitless injector, a FID and an ODP 3 Olfactory Detector Port (Gerstel, Mulheim, Germany). At the column outlet, the eluate was split 1:1 to simultaneously detect volatile compounds by FID and sniffing. Samples (4 µL) were separated on a 60 m × 0.32 mm i.d., 0.25 µm film thickness DB-5 column (J & W Scientific) according to the same temperature programmed as reported for the GC-MS analysis. Helium was used as a carrier gas at the flow rate of 1.5 mL min⁻¹. Injector and FID temperature was 250 °C; hydrogen, air and nitrogen (make up) flow rates were 30, 450 and 30 mL min⁻¹ respectively. The sniffing was performed by three trained panellists, who presented satisfactory sensitivity and reproducibility and agreement with one another. For each GC-O analysis, sniffing was divided into three sessions of 20 min and panellists were asked to characterise the detectable aroma with a freely chosen descriptor.

The FD factors of the odour-active compounds were determined by AEDA [32]. Each bread sample (200 g) was immediately frozen with liquid nitrogen and later homogenized with anhydrous sodium sulphate (200 g) in a commercial blender (Moulinette, Numberg, Germany). The powder was extracted with dichloromethane (1 L) for 8 h at 40 °C using a Soxhlet extractor. The extract was concentrated to ~100 mL by distilling off the solvent. The distillate was extracted with an aqueous sodium carbonate (0.5 mol/L, 3 × 50 mL) to remove acidic volatiles. The aqueous solution was washed with dichloromethane (50 mL), and the organic phase were combined, dried over anhydrous sodium sulphate, filtered, and concentrated to ~0.1 mL. The following series: the neutral-basic and acidic fractions were step wisely diluted with dichloromethane (1 + 1, v/v), and each dilution was analyzed by gas chromatography/olfactometry to give the flavour dilution (FD) factors of odour-active compounds.

Raw data obtained from each sensory session were averaged across panellists, and the results were studied by one-way analysis of variance (ANOVA). Duncan’s multiple range tests were used to determine significance among results. Partial least square discriminant analysis (PLS-DA) and PLS-regression coefficient were chosen as an exploratory technique to describe and summarise data by grouping variables that are correlated. The multivariate statistical analyses were performed with the SIMCA-P software (v. 12.0, Umetric, Umeå, Sweden) [33] PLS regression is specifically designed to determine relationships which exist between blocks of dependent (Y, bread types) and independent (X, volatiles) variables by seeking underlying factors common to both sets of variables [24].