Sensory evaluation as a tool in assessing the quality of new fermented products

Ten starter cultures of lactic acid bacteria were used to ferment five mixtures of milk and pea protein (0%, 10%, 20%, 30% and 40% of pea) to select the cocktail that can lead to products similar to traditional yogurt. Product quality evaluation was performed by comparing the sensory profile of 49 formulated products with the profile of a milk fermented by commercial lactic ferments. The sensory profiles were analyzed by means of three-way ANOVAs and a principal component analysis (PCA). Substitution of cow milk protein with 40% of pea proteins reduce starter cultures effects and decrease product quality. In contrast, until 30% of pea protein, starter cultures show positive and negative effects. For example, products fermented by Streptococcus thermophilus + Lactobacillus acidophilus with 30% pea protein have positive characters like creamy and smooth, but Lactobacillus delbrueckii subsp. Bulgaricus + Lactobacillus rhamnosus caused bad quality and negative characters like bitter and astringent even with 100% cow milk.


INTRODUCTION
The increase in global demand of animal proteins in recent years may become a major issue in a close future. In order to initiate the reduction in the use of animal proteins in European diet, it might be interesting to combine dairy proteins with vegetable proteins in products already known by consumers.
Tu, et al. [1] have shown that consumers can accept dairy-like products combining cow milk and soybean proteins if the ratio of soybean protein does not excess 50%. However, soybean might not be the best substitution protein as consumers tend to have negative attitudes towards this source of protein [1,2]. Zare, et al. [3] have suggested lentil flour as an alternative protein substitution. They showed that supplementation with 1-3% of lentil flour did not affect much sensory properties and overall acceptance compared to traditional yogurt. However, this rate of substitution is not high enough to initiate In this paper, we propose pea protein as a good substitute for cow milk protein in dairy products, because of its balanced amino acid profile, low level of allergy, functional properties and availability at an affordable price. Pea could be a better substitute for milk protein than soybean because of its high digestibility level, the absence of phytoestrogens and its environmentally friendly and local agriculture.
Although pea protein has positive characteristics and has been used in sports foods and in meat based products, its use is almost absent in fermented products due to its intense flavor and odor.
The objective of this research was to determine the combination of starter culture and pea concentration that will give a dairy product close to traditional yogurt. A standard descriptive analysis approach was used to compare the sensory profiles of fermented products obtained by fermenting five ratios of cow/pea milks with 10 starters with that of a cow milk yogurt fermented with a commercial milk ferment. Figure 1 illustrates the yogurt preparation.

Panels
The panel was composed of 10 women between the ages of 22 and 50. All panelists were recruited amongst the students and staff of AgroSup Dijon, France. They attended 10 one-hour training sessions, one per week one session for selection, two sessions for vocabulary generation, four sessions for training and three sessions for panel performance evaluation).
After that, the trained panelists evaluated in duplicate the 50 products (two one-hour sessions each week for 4 weeks).

Selection
One selection session was carried out for testing the panelists' ability to detect tastes (i.e., bitter, acid) and odors (butter, herbs, peas, and earth) on pea "yogurt" as well as their verbal fluency and ability to describe products. The 12 panelists with the highest detection performance, verbal fluency and ability to describe a product were selected. Among those 12 panelists, two abandoned the panel due to availability problem.

Generation of attributes
Attribute generation was conducted in two sessions. In the first session, panelists were asked to describe with their own words five samples selected among the 50 possible products so as to span as much as possible the product sensory space. Generated attributes were compiled to form a preliminary list. In the second session, panelists were presented five new products which they had not been exposed to before, and were asked to rate every attributes on a 6-point intensity scale (from 0 to 5) using the preliminary list of attributes. Panelists were free to add attributes to the list if necessary. A reduction of the list of attributes was then performed following the ISO 11035:1994 standard.

Training procedure
During training, panelists agreed upon definitions, references and procedures for each attribute and were trained to rank different water and yogurt solutions containing substances that give the required attributes (e.g. acid lactic for the acid attribute or caffeine for the bitter attribute). Finally, 10 new products were presented in duplicate to determine whether the panel was homogeneous, discriminant and repeatable.

Final profiling
The final profiling consisted of eight one-hour sessions (two sessions a week). Fifty products were evaluated in duplicate on a structured interval scale going from 1 (low) to 10 (high).

Panel performance
A three-way ANOVA was carried out for each attribute with the following model: Both assessor and product were considered as fixed factors. When a significant product  assessor interaction was found, a principal component analysis (PCA) was performed to evaluate the consensus between panelists. Data were analyzed using SPAD version 7.4.

Product description
Intensity scores obtained for each attribute were averaged across repetition and submitted to threeway ANOVAs with assessor, starter culture and pea concentration as within subject factors. Assessor was considered as a random factor and both starter culture and pea concentration as fixed factors. Attributes with a significant effect of either starter culture or pea concentration were then submitted to a normalized principal component analysis (PCA) and a hierarchical cluster analysis (HCA). ANOVA were performed using SAS 9.3, and PCA and HCA with the SPAD 7.4.

Attributes generated by the panels
To describe the products, the panelists used 13 attributes including four of texture (astringent, fluid, smooth, creamy) three of taste (sweet, bitter, acid) and six of aroma (vinegar, earth, vegetable, smoked, dairy, pea). The 13 attributes were classified as positive or negative attributes based both on the literature [6] and on a preliminary study (Table 2).

Panel performance
The product effect was significant for all the attributes at the 5% level. Therefore, the panelists were able to discriminate between the 50 products.
The repetition effect was significant at the 5% level for six descriptors (fluid, creamy, smooth, astringent, bitter and acid). This repetition effect can however be due in part to differences in the products rather than in the panelists. A significant interaction assessor x product was found for all attributes. However, the PCA performed on each descriptor with assessors as variables, showed a good consensus between assessors except for the attribute bitter.

Product description and comparison with the standard 3.3.1. ANOVA: product description
The three-way ANOVA (    The HCA performed on the first two PCA dimensions showed that the 50 products could be divided into four classes ( Table 4). The first class includes eight products, seven of them without pea protein. It has been described with positive attributes close to the attributes of traditional yogurts usually consumed by the panel. Tu, et al. [6]. Moreover, products fermented by traditional starter culture Streptococcus thermophilus + Lactobacillus delbrueckii subsp. Bulgaricus have good evaluation (neither acid nor astringent). The second class includes 10 products; two of them have been described with negative descriptors as bitter, astringent and acid despite the absence of pea protein.
This may be caused in "I" starter culture by the high ability of acidification of Lactobacillus rhamnosus that could decrease sensory characteristics, [7], or by the high esterase activities of the two strains in "E" starter culture Lactobacillus delbrueckii subsp.
Bulgaricus + Lactobacillus helveticus [8]. The third group includes promising products, which were not associated with negative characteristics such as astringent, acid or bitter and have received high scores for positive descriptors like sweet, smooth and creamy. The metabolic activity of some microorganisms such as Lactobacillus acidophilus [9], and Lactobacillus casei [10], results in production of flavor, and aroma that cause good organoleptic properties. These organoleptic properties could be cumulated to those of traditional strains used in fermented products (Streptococcus thermophilus and Lactobacillus delbrueckii subsp. Bulgaricus), when they were used together as starter cultures. The main quality caveat of this group of products is the presence of pea-derived aromas (pea, vegetal and earth) which might be reduced using aromatization processes. The fourth class includes products with 40% of pea protein, which are characterized by high intensity of pea aromas. With this level of pea protein, no starter culture was able to decrease the negative characteristics of pea protein.