风味作为食品品质与消费者接受度的核心要素,其不仅涉及物质层面的化学组成,更涵盖了生理感知与心理愉悦的跨学科交互[1-2]。但是风味形成与感知机制的复杂性与跨学科性,导致风味研究的难度较大。因此,即使是“同一加工条件”,在不同的产地、成熟期、品种的原料中也会表现出迥异的风味特征。因此,风味科学的核心挑战经过几十年的演化,已从“找到某个关键风味成分”转向“在复杂体系中建立可解释、可验证的化学组分与人类感觉映射之间的因果联系”[3]。需要特别指出的是,广义的“感官”(sensory)与“风味”(flavor)是一个由嗅觉(气味)、味觉(滋味)、视觉(色泽)、触觉(质构)等多维度构成的感知复合体。然而,受限于目前解析技术瓶颈,当前绝大多数的风味定性定量工作与高通量分析技术仍高度聚焦于挥发性气味活性物质与部分非挥发性滋味成分。相较之下,关于色泽、质构等维度的分子级关联研究相对匮乏。因此,为保证论述的深度与聚焦性,本文所探讨的“感官组学”与“风味组学”及其融合范式,将主要限定在狭义的化学感知范畴内,即以化学感知中气味为主要路径,以物质解析及感知机制为核心展开讨论。

为了更好地解决上述问题,感官组学(sensomics)与风味组学(flavoromics)两种风味研究范式逐渐发展为目前风味研究中的核心研究方法[4],前者通过感官驱动的标准化风味研究流程与受到培训的感官评价小组,研究食品体系中与风味直接相关的风味活性成分及其对食品感官属性的影响;后者则侧重于利用代谢组学的高通量分析技术,对食品中的风味物质进行全面定性定量分析,基于数据驱动建立食品中的风味组分与感官属性或消费者偏好性甚至情绪感知之间的可能联系[5-7]。本文以风味科学研究中两种常见的范式——感官组学与风味组学为主线,对近年来利用两种方法开展的相关研究进行归纳总结,着重介绍两种方法的优缺点,并在此基础上提出风味科学研究的未来新趋势,进一步探讨风味组学与感官组学融合的未来新路径,为食品风味相关研究发展提出新的思考。

1 基于关键风味成分解析和感官重组的感官组学研究范式

感官组学(sensomics)是由德国慕尼黑工业大学科学家Peter Schieberle于2007年正式提出,也称“分子感官科学”(molecular sensory science)[8]。该方法强调以感官为导向,整合气相色谱-质谱-嗅觉测定法(gas chromatography-olfactometry-mass spectrometry,GC-O-MS)、香气提取稀释分析(aroma extract dilution analysis, AEDA)、气味活性值(odor activity value,OAV)、香气重组与缺失实验等技术,旨在鉴定并量化真正具有感官贡献的关键香气活性化合物,即能实际影响整体香气感知的分子[9-10]。

感官组学研究范式主要分成3个步骤,见图1。首先是关键风味活性组分的定性,其次是关键风味活性成分的精确定量,最后是通过缺失与重组实验来对上述确定的风味活性成分进行验证[11]。受到训练的评价员对于样品的感觉测量贯穿整个研究过程,首先是风味活性成分的定性阶段,需要结合嗅闻或滋味品尝实验;其次是精确定量环节,需要结合化合物的气味或滋味阈值(评价员测量)计算其风味活性值(odor activity value,或者taste active value,TAV);最后在重组验证阶段;需要结合经培训的专家感官评价小组开展定量描述分析(quantitative descriptive analysis,QDA)实验以及基于差别检测的缺失实验(例如,三点检验、二三检验等)[12]。

图1 感官组学研究范式

Fig.1 Sensomics research paradigm

感官组学以“化学分子能被人的感官所感知”为筛选目标,主要目的是建立从化学成分的定性定量分析到专家感官属性强度之间的因果链条,其核心在于在每一步的分析过程中都融入感官测量,把“测得到的风味化合物仪器信号”转化为“闻得到/尝得到的感官贡献”,从而建立感官相关的因果联系[13]。目前基于感官组学的方法,已经有多类食品体系的研究,包括但不限于果蔬/果汁(如哈密瓜/甜瓜果汁等)、酒类(葡萄酒/啤酒等)、乳制品(牛奶、酸奶、奶酪等)、发酵水产/发酵调味品等[14]。该方法能够在这些食品中识别与验证决定典型香气表型的关键气味活性化合物,通过缺失/重组实验区分“关键”与“次要”贡献分子并评估不同化合物之间的协同/掩蔽效应[15-16]。例如,Wagner等[17]利用分子感官科学理念,结合 AEDA 筛选、稳定同位素稀释分析(stable isotope dilution analysis, SIDA)精确定量及OAV计算,系统解析了生甘草中的关键致香成分,并通过风味重组实验成功复原了其特征香气,从而确证了关键风味物质的准确性。关于感官组学使用过程中的具体步骤,以及实施过程中需要注意的问题,可以参考团队已发表综述[18]。

感官组学也存在研究方法的局限性,常见如下:在风味活性成分定性过程中常用的方法,例如GC-O等不可避免地带有主观性和条件依赖性[19]。例如,评价小组的训练水平、嗅闻疲劳程度、记录方式以及阈值标准等因素,均会影响对气味有无及强弱的判断。虽然为了进一步提升GC-O的量化效果,研究者常采用稀释分析法,如香气提取稀释分析(AEDA)等[20],然而这类基于稀释因子的强度排名并不能完全等同于化合物对整体风味的真实贡献。其中一个主要原因是,色谱柱上可能会发生“峰共洗脱”现象,即多个化合物同时从色谱柱洗脱并通过嗅闻端口,这会显著干扰嗅闻判断,导致对单个化合物的感知失真。此外,不同化合物在嗅闻过程中的释放行为并不一致,且嗅觉的剂量-反应关系也因物质而异(例如,某些化合物在低浓度下即可被显著感知,而另一些则需要更高浓度才能产生明显的气味效应)[21-22]。这些复杂的因素共同作用,可能导致基于稀释因子的排序结果与化合物在实际风味中的真实贡献产生偏差。

其次,在风味活性成分定量阶段,感官组学强调定量方法的准确性和可重复性,因此目前挥发性组分气质分析方法的黄金标准是溶剂辅助蒸发(solvent assisted flavor evaporation,SAFE)萃取结合稳定同位素稀释分析(SIDA)[23]。SIDA 校正法能够有效抵消前处理损失、基质抑制和仪器漂移的影响[24]。但由于SAFE方法也是对样品中风味组分进行萃取浓缩的方法,其也存在萃取过程中风味组分含量的偏移,并不是食品上方顶空气体的真实含量,只是一种相对准确的并稳定的定量方法,并且SAFE过程中稳定同位素标品的损失量较大,导致该方法成本高昂,因此SAFE是否作为金标准仍是未来值得深入探讨的话题[25-26]。此外,在精确定量数据的基础上,气味活性值(odor activity value, OAV),即通过化合物浓度与嗅觉阈值的比值计算而来,常被用来连接“浓度”与“潜在嗅觉贡献”[27-28]。OAV虽然有助于解释为何某些含量不高的成分仍可能主导风味[29-30],然而必须强调的是,嗅觉阈值并非一个恒定不变的参数。首先,基质的脂相/水相分配、结合与释放行为都会改变目标化合物在气相中的有效浓度,从而影响其嗅闻强度[31]。其次,食品体系中的酸度、盐度与甜味背景等因素也会直接作用于嗅觉受体,进而改变对特定化合物的感知阈值。此外,风味化合物之间的协同、掩蔽与交互作用更为复杂,可能导致“高OAV不等于高贡献”或“低OAV不等于无贡献”的情况[32-33]。因此,OAV更适合作为指导性线索或优先级排序工具,而非直接的最终结论。

最后,缺失重组实验环节的可重复性高度依赖操作细节:基质配方与载体选择、溶剂挥发行为、复配顺序、释放条件(温度、搅拌、顶空体积)、嗅闻方式与评价任务(QDA/差异检验)都会影响结论的重复性。此外,除了前述挑战外,感官组学方法本身耗时长且成本高,导致数据积累缓慢。

感官组学方法的核心优势在于其以感官响应为终点,并提供可解释的证据,侧重于因果关系的验证[34-35]。其局限性在于覆盖面和测量通量通常有限,难以对复杂的化学空间进行全局刻画。鉴于该方法的局限性,特别是在处理复杂体系时的不足,研究人员将目光投向了风味组学的分析策略。风味组学通过整合高维度化学数据和先进的数据分析技术,能够有效弥补感官组学在覆盖面和全局性描述上的不足[36]。

2 基于高通量指纹与数据驱动的风味组学研究范式

风味组学(flavoromics)是在2011年由美国风味化学家Gary Reineccius提出的[37]。该方法借鉴了现代代谢组学的思想,通过先进分离分析技术(GC-MS或HPLC-MS等)获取大量风味成分信息,再结合多元统计或机器学习建模,挖掘与感官属性(“果香”、“奶香”等)相关联的关键差异性化合物(differential compounds)。风味组学的目标在于基于风味组分对食品进行分类、鉴别与快速表征,而非严格验证风味物质与感官之间的因果联系[38]。

风味组学的研究方法也分成3个主要步骤。首先是基于仪器分析对食品中风味组分物质基础开展定性与定量研究;其次是研究专家评价员或消费者对食品感官属性或感知响应的量化;最后是利用统计学方法建立输入数据(食品中风味物质基础)与输出数据(人的感官感知响应)之间的联系并进行模型验证[6,39],见图2。

图2 风味组学研究范式

Fig.2 Flavoromics research paradigm

风味组学的核心思想在于构建“非靶向、广覆盖”的化学指纹,并通过数据驱动进行挖掘[6]。其方法论优势体现为不预设目标物,而是借助高分离度与高通量的仪器平台(GC×GC-TOF MS联合LC-MS用于挥发物及非挥发性代谢物全景捕捉等),以获取尽可能完整的分子特征矩阵,从而在复杂体系中发现未预期的差异信号[40-41]。相较于仅关注少数已知关键物的靶向策略,全景分析方法因其对分子特征的广谱覆盖,更能有效回答“整体异同及差异所在”这类整体性问题。具体而言,其常见的应用价值体现在多个方面:一是可用于建立产地、品种或工艺差异的可追溯指纹分析,实现产品的来源识别与真伪鉴别;例如,De Tdedo等[42]利用判别分析(DA)方法,基于25个全球产地烘焙咖啡样品的挥发性化合物组成,成功构建了可区分咖啡大陆级(非洲、亚洲、中美洲、南美洲)及国家级(尤其是巴西)产地的统计模型,从而为咖啡原产地真实性与质量控制提供了可靠的化学计量学工具。二是能够监测货架期劣变等过程中风味化合物的整体变化趋势,揭示其动态变化规律;例如,Corbo等[43]研究了天然挥发性化合物与贮藏温度协同作用对新鲜切片苹果微生物稳定性的影响,通过实验证实添加低浓度己醛可显著延长货架期,增强颜色稳定性,并在不同温度下仍有效抑制微生物生长。三是在配方替换与降低成本等应用场景中,可快速识别潜在风险区段,预判对风味质量的影响。例如,Peng等[44]开发了一种结合挥发性代谢组学与机器学习的可泛化方法,成功实现了对武夷岩茶核心产区与非核心产区来源的高精度鉴别,并筛选出20种关键差异挥发物,揭示了“岩韵”风味的产地特异性化学基础。然而,风味组学全景指纹的价值能否充分有意义,很大程度上取决于常被忽略的样品预处理与数据质量控制,这往往是决定其结论可信度的关键节点[45]。

风味组学侧重于高通量的数据驱动,但是在研究方法上也存在一定局限。总体而言,这些局限主要来自仪器分析、感官评价与数据建模三个层面。仪器层面,首先,由于谱库覆盖不全以及食品样品成分复杂、易发生化合物共洗脱,风味物质的定性注释可能出现误判,甚至组分定性错误的情况[46]。其次,样品前处理与进样条件(如富集方式选择、高温进样)可能带来选择性偏差,并在某些情况下生成并非样品原本存在的分析伪影[47-48]。再次,目前大部分研究采用的顶空萃取定量方式,容易受到基质效应影响,导致不同批次、不同研究之间的数据难以直接比较。最后,仪器通常获得的是相对静态的挥发物谱,而人体真实感受到的风味大部分来自口腔加工过程,其与口腔加工、温度/湿度、基质释放与呼吸等因素形成的动态暴露等过程均相关,因此“分析对象”与“感受过程”存在一定错位[49-53]。

在感官层面,常用的QDA以及CATA/RATA等感官评价方法多为静态或终点评价,难以刻画风味从入口、达到峰值到余味的动态变化,也不易区分鼻前嗅觉(闻)与鼻后嗅觉(口腔加工与吞咽)的差异。此外,评价人员个体差异、训练程度不同以及术语定义不统一,会降低不同研究结果之间的可重复性与可比性[54-55]。与此同时,食品体系中普遍存在基质效应以及组分间的协同或掩蔽作用,使得单个化合物浓度或OAV与实际感知强度往往不呈简单线性关系,也进一步限制了评价结果的准确性。

在建模层面,PCA、PLS-DA、随机森林等统计或机器学习方法在风味组学中应用广泛,但多数侧重于发现“相关性”,并不等同于发现“因果关系”[56-57]。此外,目前食品体系数据大多属于高维小样本的典型数据结构,模型容易过拟合,且对峰提取、缺失值处理、归一化与尺度变换等预处理步骤高度敏感,导致“关键标志物”不稳定。同时,风味感知中的非线性、组分交互与时间动态特征难以被静态模型充分表达。因此,如需形成更可靠的机制解释,仍应结合组分重组、遗漏实验等靶向验证手段完成证据闭环[58-59]。

为了深入阐明风味形成的机制,风味组学的研究目标超越了单纯的成分鉴定,进而通过整合多组学数据与过程信息,探究作用机制[60]。具体而言,这意味着将样品中的挥发性与非挥发性等风味物质特征以及其他相关数据进行关联,例如与转录组学(基因表达谱)、代谢组学(代谢产物指纹)、微生物组学(微生物群落结构)等多组学信息以及关键工艺参数(温度、pH值、氧暴露程度、加热工艺等)进行关联分析。通过这种多维度关联,可以追溯到驱动风味形成的可能生物化学或食品化学的路径,例如,Yang等[61]研究了低盐浓度(11%、14%、17%)对豆瓣酱发酵过程中理化性质、风味物质、生物胺及微生物群落动态的影响,发现减盐能够促进氨基酸氮、游离氨基酸和多数挥发性风味物质积累、提升风味复杂度。此外,有研究比较了半干型客家黄酒(SHRW)与传统甜型客家黄酒(TSRW)在发酵过程中理化性质、微生物群落及风味代谢物的动态变化规律,并通过相关性分析揭示了关键微生物与特征风味物质间的潜在关联,为优化客家黄酒发酵工艺和品质调控提供了理论依据与数据支撑[62-64]。然而,必须强调的是,相关性分析所揭示的关联并非等同于因果关系。为实现对风味机制的严谨验证和因果关系的建立,机制推断最终必须通过受控的实验验证来完成闭环。这包括开展靶向定量分析、关键成分的添加或缺失实验,以及稳定同位素标记底物在模拟体系中验证等实验,以验证该特定因素对风味形成的直接影响[3,65]。

总体而言,风味组学的优势在于其全面的覆盖范围、强大的未知化合物发现能力以及对复杂风味体系的系统性指纹化分析。然而,其局限性在于,统计学上的差异特征并不总是直接等同于实际的风味贡献[66]。因此,风味组学的研究成果必须与专业的感官评估证据以及前述的因果验证框架紧密结合,才能实现从单纯的差异特征联系的建立到真正决定产品风味的关键因素的识别与理解(见表1)。

表1 感官组学与风味组学研究范式对比

Tab.1 Comparison of research paradigms between sensomics and flavoromics

对比维度感官组学风味组学核心逻辑因果验证(自上而下)相关性挖掘(自下而上)方法学特征感官评价小组(QDA)、稀释分析(AEDA)、重组与缺失验证高通量全景质谱扫描、多元统计与机器学习建模主要优势精准确证关键呈香物质,机制解释力强覆盖面广,可发现未知/痕量标志物,无须预设目标物主要局限通量低、成本高;难以捕捉基质效应与非线性协同作用易过拟合;静态数据难匹配动态感知;缺乏因果证明典型应用场景特征风味重塑、关键异味溯源真伪与产地鉴别、货架期与工艺监测未来融合方向提供感官反馈与因果验证基准提供高维数据输入与复杂网络挖掘算法

3 风味范式融合的挑战与趋势

传统的风味解析常以单一理化指标或少数挥发物含量为导向,在复杂基质与动态加工中往往面临3大瓶颈:其一,仅凭含量或单变量难以解释整体风味,常出现“含量高却不显著”“含量低却关键”的现象,本质上源于化合物阈值、风味组分协同作用及风味组分与食品大分子基质之间的释放差异[67-68];其二,跨批次、跨平台与跨实验室的可比性不足,样品前处理、顶空平衡条件、色谱柱与离子源、数据处理参数均会导致显著偏差,削弱结论的稳健性[69-70];其三,从“发现”到“验证”的链条常被割裂——非靶向筛到的候选标志物缺乏感官因果检验,靶向锁定的关键活性物又可能忽略背景复杂化学空间与基质调控,导致研发决策依赖经验而难以复现[71](见图3)。

图3 传统风味解析方法的局限性

Fig.3 Limitations of traditional flavor analysis methods

此外,针对“动态感知机制”这一难题,目前研究中常采用实时监测仪器与动态感官评价相结合的方法来应对。在仪器分析方面,逐渐由传统的静态取样转向在线动态监测(在线质谱或实时电子鼻等),以获取风味物质在加工或咀嚼过程中的时间释放曲线。在感官评价方面,研究者广泛引入了时间-强度(TI)和感觉时序支配(TDS)等动态感官评价方法,用于记录特征风味随时间消长的过程[72]。通过将这两者的时间序列数据进行相关性分析,可以初步建立“化学释放-主观感知”的动态对应关系,从而有效弥补传统静态分析的不足。

为了打破“仪器检测的化学物质”与“人类感知的风味体验”之间难以精准对应的局限,感官组学(侧重因果律确立)、风味组学(侧重高通量挖掘)与人工智能(AI)技术的深度融合正成为风味解析的新范式(见图4)。在这一融合范式中,AI技术不仅是数据处理的工具,更是解决高维复杂系统非线性映射的核心。具体而言,AI主要解决了两类核心问题:一是多组学数据的深度整合,通过深度学习使多模态数据充分融合,有效消除跨平台仪器的批次效应[73];二是关键风味分子与感官属性的精准预测,利用随机森林(RF)、支持向量回归(SVR)等机器学习模型,能够量化风味物质之间的协同效应,并在未进行大规模感官评价的情况下,基于分子结构和感官评价数据预测食品的感官描述词与喜好度[74]。

图4 范式融合综合性示意

Fig.4 Comprehensive schematic diagram of paradigm integration

目前的风味范式融合研究已在多类复杂食品体系中展现出巨大潜力[75]。以酱香型白酒的“空杯留香”为例,研究者首先利用感官组学锁定关键感官特征,随后将感官数据与非靶向代谢组学数据进行多变量交叉验证,结合机器学习模型,成功锁定了对风味贡献显著且阈值极低的痕量含硫或吡嗪类化合物,将模糊感官体验转化为可调控工艺参数[76]。此外,在牛肉腐败检测中,已有研究利用AI算法整合多模态传感数据和挥发性化合物、低场核磁等数据,精确监控牛肉腐败过程并预测其贮藏时间[77];在初榨橄榄油品质鉴定中,研究将质谱、光谱等多源数据与专家感官评分深度融合,依托偏最小二乘判别(PLS-DA)模型,显著提升了对发霉、酸败等关键感官缺陷的分类准确率,有效突破了单一分析仪器的局限[78]。这些典型案例有力证明,多组学与人工智能融合的新范式,能够建立“特定加工参数-化学动态变化-感官影响”之间的稳健因果联系,为食品精准定向调香与工艺优化提供了坚实的科学基础。

4 结论与展望

风味科学正经历着深刻的范式变革。它正从过去侧重于单一化学成分鉴定的描述性研究,逐步转向强调成分与感官多维交互的预测性科学。这种转变的核心在于,风味研究不再局限于单一维度的标志物定量,而是将其视为化学信号与感官感知在多维时空下的系统集成,旨在更深入地刻画分子组成与感知体验之间的内在逻辑。

在研究路径上,感官组学坚持“感官引导下的定性定量与重组验证”这一经典范式,通过为关键活性分子建立确凿的因果链条,提供了深入的机制见解,但往往受限于低通量与高昂的实验成本;而风味组学则利用先进的质谱分析仪器配合数据统计工具,能够在大规模样本信息中快速提取化学指纹信息,实现全景式的成分扫描,但其数据输出往往面临缺乏直观感官解释性的瓶颈。针对上述两种路径各自的局限,加强二者的深度融合成为未来风味科学发展的一条可行道路。通过实现优势互补,这种融合不仅能打破单一学科的认知边界,更能构建出兼具优秀机制解释能力与预测稳定性的风味解析新范式。

然而,面向未来复杂食品体系的定向创制需求,这种融合仍需在深度与广度上进一步拓展。未来感官组学与风味组学的融合应在3个方向寻求具体实施与突破。

1)构建标准化与本土化的多模态风味数据库。目前跨组学融合的主要阻碍在于数据稀疏与底层标准不一。特别是在中式传统饮食体系中,诸如“醇厚”、“回甘”等复杂感官属性尚未完全实现分子级别的标准化定义。在具体实施中,应率先建立统一的样品前处理协议与感官评价量表,整合色谱质谱高通量数据、受体结合靶点信息以及规范化的感官描述词汇。通过构建结构化的“风味知识图谱”,可以为后续的数据挖掘提供坚实的数据基建,使复杂感官经验的数字化复现成为可能。

2)深化从“静态组成”向“时空动态感知”的机理解析。食品风味并非分子的简单叠加,而是受食品大分子基质动态调控的复杂释放过程。未来的融合范式将更多地引入动态监测技术,同时结合动态感官评价模型进行深入分析。通过解析风味物质在口腔加工、咀嚼及吞咽过程中的时空释放动力学,研究者可以更真实地还原人类的感官体验。这种方法能够明确食品基质(蛋白、多糖等)如何影响风味的释放速率,从而为精准调控产品风味的持久度提供技术支持。

3)依托人工智能实现风味的数字化闭环与定向创制。随着机器学习与深度神经网络的介入,风味解析正从“事后解释”走向“事前预测”。通过AI强大的高维非线性计算能力,科研人员可以高效破译化学组分间的协同或掩蔽效应。在实际应用中,这不仅能通过算法模拟预测新口味的市场接受度,还能为当前行业热点的植物基食品、减糖减盐等健康食品提供精准的“风味补偿”方案。通过这种“配方逆向工程”,最终能够实现从分子设计到终端产品的数字化研发闭环,显著提升食品创制的预见性与操作性。

参考文献：

[1] SERGI G, BANO G, PIZZATO S, et al. Taste loss in the elderly: possible implications for dietary habits[J]. Critical Reviews in Food Science and Nutrition, 2017, 57(17): 3684-3689.

[2] TEES R C. Review of the organization of behavior: a neuropsychological theory[J]. Psychologie Canadienne, 2003, 44(1): 74-76.

[3] KLEE H J. Improving the flavor of fresh fruits: geno-mics, biochemistry, and biotechnology[J]. New Phytologist, 2010, 187(1): 44-56.

[4] SONG Z W, DU H, ZHANG Y, et al. Unraveling core functional microbiota in traditional solid-state fermentation by high-throughput amplicons and metatranscriptomics sequencing[J]. Frontiers in Microbiology, 2017, 8: 1294.

[5] ZHANG B H, YANG Y B, SONG F H, et al. Decoding Baijiu flavor complexity: integrating dynamic sensory analysis, high-resolution metabolomics, and advanced mass spectrometry[J]. Food Research International, 2025, 221: 117371.

[6] SCHREURS M, PIAMPONGSANT S, RONCORONI M, et al. Predicting and improving complex beer flavor through machine learning[J]. Nature Communications, 2024, 15: 2368.

[7] KEAWKIM K, LORJAROENPHON Y, VANGNAI K, et al. Metabolite-flavor profile, phenolic content, and antioxidant activity changes in Sacha Inchi (Plukenetia volubilis L.) seeds during germination[J]. Foods, 2021, 10(10): 2476.

[8] STEINHAUS P, SCHIEBERLE P. Characterization of the key aroma compounds in soy sauce using approaches of molecular sensory science[J]. Journal of Agricultural and Food Chemistry, 2007, 55(15): 6262-6269.

[9] VRZAL T, OLOVSK J. Sensomics-basic principles and practice[J]. Kvasny Prumysl, 2019, 65(5): 166-173.

[10] HOLT S, MIKS M H, DE CARVALHO B T, et al. The molecular biology of fruity and floral aromas in beer and other alcoholic beverages[J]. FEMS Microbiology Reviews, 2019, 43(3): 193-222.

[11] SUN J Y, LI Q Y, LUO S Q, et al. Characterization of key aroma compounds in Meilanchun sesame flavor style Baijiu by application of aroma extract dilution analysis, quantitative measurements, aroma recombination, and omission/addition experiments[J]. RSC Advances, 2018, 8(42): 23757-23767.

[12] ROTONDI A, BENDINI A, CERRETANI L, et al. Effect of olive ripening degree on the oxidative stability and organoleptic properties of cv. Nostrana di Brisighella extra virgin olive oil[J]. Journal of Agricultural and Food Chemistry, 2004, 52(11): 3649-3654.

[13] STAROWICZ M. Analysis of volatiles in food products[J]. Separations, 2021, 8(9): 157.

[14] MAVANI N R, ALI J M, OTHMAN S, et al. Application of artificial intelligence in food industry: a guideline[J]. Food Engineering Reviews, 2022, 14(1): 134-175.

[15] LI X H, ZHANG C, LI S M, et al. Unveiling key aroma-active compounds of two distinct aroma-types oriental melon: an integrated sensomics approach utilizing GC×GC-O-TOF-MS, GC-MS, aroma recombination and omission experiments[J]. Food Chemistry: X, 2025, 28: 102517.

[16] WANG J, YI X, HUANG M Q, et al. Studies on the key odorants in Maopu buckwheat finished Baijiu and the effect of Tartary buckwheat extract on its flavor[J]. LWT-Food Science and Technology, 2022, 154: 112650.

[17] WAGNER J, GRANVOGL M, SCHIEBERLE P. Charac-terization of the key aroma compounds in raw licorice (Glycyrrhiza glabra L.) by means of molecular sensory science[J]. Journal of Agricultural and Food Chemistry, 2016, 64(44): 8388-8396.

[18] HAN H Y, BLANK I, XU K L, et al. From molecule identification to sensory prediction: approaches, applications, and perspectives of sensomics and flavoromics in volatile flavor compound research[J]. Journal of Agricultural and Food Chemistry, 2026: acs.jafc.5c13311.

[19] SHANG L, LIU C J, TOMIURA Y, et al. Machine-learning-based olfactometer: prediction of odor perception from physicochemical features of odorant molecules[J]. Analytical Chemistry, 2017, 89(22): 11999-12005.

[20] SCHIRACK A V, DRAKE M A, SANDERS T H, et al. Characterization of aroma-active compounds in microwave blanched peanuts[J]. Journal of Food Science, 2006, 71(9): C513-C520.

[21] MCCLINTOCK T S, WANG Q, SENGOKU T, et al. Mixture and concentration effects on odorant receptor response patterns in vivo[J]. Chemical Senses, 2020, 45(6): 429-438.

[22] PELZ D, ROESKE T, SYED Z, et al. The molecular receptive range of an olfactory receptor in vivo(Drosop-hila melanogaster Or22a)[J]. Journal of Neurobiology, 2006, 66(14): 1544-1563.

[23] ZHU W Q, CADWALLADER K R. Streamlined approach for careful and exhaustive aroma characterization of aged distilled liquors[J]. Food Chemistry: X, 2019, 3: 100038.

[24] TADESSE F T, JEMAL Y, ABEBE H. Effect of green coffee processing methods and roasting temperatures on physical and cup quality of Sidama Coffee, southern Ethiopia[J]. Journal of Nutritional Ecology and Food Research, 2016, 3(1): 44-50.

[25] ZHANG Y F, STÖPPELMANN F, ZHU L, et al. A comparative study on flavor trapping techniques from the viewpoint of odorants of hot-pressed rapeseed oil[J]. Food Chemistry, 2023, 426: 136617.

[26] LIU M Y, ZHOU J, QIN J K, et al. Evaluation of high vacuum flavor extraction device as a novel technique for the extraction of volatile compounds[J]. Foods, 2024, 13(19): 3206.

[27] BI S, XU X X, LUO D S, et al. Characterization of key aroma compounds in raw and roasted peas (Pisum sativum L.) by application of instrumental and sensory techniques[J]. Journal of Agricultural and Food Chemi-stry, 2020, 68(9): 2718-2727.

[28] FRAUENDORFER F, SCHIEBERLE P. Identification of the key aroma compounds in cocoa powder based on molecular sensory correlations[J]. Journal of Agricultural and Food Chemistry, 2006, 54(15): 5521-5529.

[29] BUCK N, GOBLIRSCH T, BEAUCHAMP J, et al. Key aroma compounds in two Bavarian gins[J]. Applied Sciences, 2020, 10(20): 7269.

[30] SAM F E, MA T Z, LIANG Y H, et al. Comparison between membrane and thermal dealcoholization me-thods: their impact on the chemical parameters, volatile composition, and sensory characteristics of wines[J]. Membranes, 2021, 11(12): 957.

[31] FRANK D, APPELQVIST I, PIYASIRI U, et al. Proton transfer reaction mass spectrometry and time intensity perceptual measurement of flavor release from lipid emulsions using trained human subjects[J]. Journal of Agricultural and Food Chemistry, 2011, 59(9): 4891-4903.

[32] ROBINSON A L, BOSS P K, SOLOMON P S, et al. Origins of grape and wine aroma. Part 2. Chemical and sensory analysis[J]. American Journal of Enology and Viticulture, 2014, 65(1): 25-42.

[33] BAX C, SIRONI S, CAPELLI L. How can odors be measured? An overview of methods and their applications[J]. Atmosphere, 2020, 11(1): 92.

[34] GRÜNBAUM D, STERN M L, LANG E W. Quantitative probing: validating causal models with quantitative domain knowledge[J]. Journal of Causal Inference, 2023, 11: 20220060.

[35] SUBBASWAMY A, CHEN B, SARIA S. A unifying causal framework for analyzing dataset shift-stable lear-ning algorithms[J]. Journal of Causal Inference, 2022, 10(1): 64-89.

[36] ZHANG X H, ZOU J L, XIONG L J, et al. Comparative characterization of flavour metabolites in meat of Qianbei brown goat, Sunit sheep, and Tan sheep using headspace solid-phase microextraction gas chromatography-mass spectrometry and relative odour activity value[J]. International Journal of Food Science and Techno-logy, 2025, 60: vvae075.

[37] CHARVE J, CHEN C, HEGEMAN A D, et al. Evaluation of instrumental methods for the untargeted analysis of chemical stimuli of orange juice flavour[J]. Flavour and Fragrance Journal, 2011, 26(6): 429-440.

[38] ZHANG S C, CHEN J N, GAO F H, et al. Foodomics as a tool for evaluating food authenticity and safety from field to Table: a review[J]. Foods, 2025, 14(1): 15.

[39] FENG T, SHUI M Z, SONG S Q, et al. Characterization of the key aroma compounds in three truffle varieties from China by flavoromics approach[J]. Molecules, 2019, 24(18): 3305.

[40] WANG Z X, CHEN L Q, YANG F, et al. High-sensitivity profiling of dipeptides in sauce-flavor Baijiu Daqu by chemical derivatization and ultrahigh-performance liquid chromatography coupled with high-resolution mass spectrometry[J]. Food Chemistry: X, 2024, 21: 101097.

[41] ZHAO Y, ZHOU J, LUO W, et al. An integrated metabolomics-flavoromics analysis reveals flavor diversity in indigenous and hybrid cattle via UHPLC-MS/MS and GC×GC-TOF-MS[J]. Food Chemistry: X, 2025,29:102801.

[42] DE TOLEDO P R A B, DE MELO M M R, PEZZA H R, et al. Discriminant analysis for unveiling the origin of roasted coffee samples: a tool for quality control of coffee related products[J]. Food Control, 2017, 73: 164-174.

[43] CORBO M R, LANCIOTTI R, GARDINI F, et al. Effects of hexanal, trans-2-hexenal, and storage tempera-ture on shelf life of fresh sliced apples[J]. Journal of Agricultural and Food Chemistry, 2000, 48(6): 2401-2408.

[44] PENG Y F, ZHENG C, GUO S, et al. Metabolomics integrated with machine learning to discriminate the geographic origin of Rougui Wuyi rock tea[J]. npj Science of Food, 2023, 7: 7.

[45] KRALJI K, STJEPANOVI T, OBRANOVI M, et al. Influence of conditioning temperature on the qua-lity, nutritional properties and volatile profile of virgin rapeseed oil[J]. Food Technology and Biotechnology, 2018, 56(4): 562-572.

[46] NICOLOTTI L, CORDERO C, CAGLIERO C, et al. Quantitative fingerprinting by headspace—two-dimensional comprehensive gas chromatography-mass spectrometry of solid matrices: some challenging aspects of the exhaustive assessment of food volatiles[J]. Analy-tica Chimica Acta, 2013, 798: 115-125.

[47] CHAUDHARY P R, JAYAPRAKASHA G K, PATIL B S. Headspace and solid-phase microextraction methods for the identification of volatile flavor compounds in citrus fruits[M]∥Instrumental Methods for the Analysis and Identification of Bioactive Molecules. Washington, DC: American Chemical Society,2014: 243-256.

[48] SPANIER A M, ST ANGELO A J, GRIMM C C, et al. Relationship of temperature to the production of lipid volatiles from beef[M]∥Lipids in Food Flavors. Washington, DC:American Chemical Society, 1994: 78-97.

[49] WILSON A D, BAIETTO M. Applications and advances in electronic-nose technologies[J]. Sensors, 2009, 9(7): 5099-5148.

[50] 田星, 穆馨怡, 邓慧琳, 等. 口腔加工对于食品风味感知及其释放影响的研究进展[J]. 食品研究与开发, 2021, 42(8): 186-191.TIAN X, MU X Y, DENG H L, et al. Advances in research on the effects of oral processing on food flavor perception and release[J]. Food Research and Development, 2021, 42(8): 186-191.

[51] 何晶, 邱丽霞, 张迪, 等. 热风干制凡纳滨对虾口腔加工过程中香气感知特性的研究[J]. 食品与发酵工业, 2024, 50(13): 223-231.HE J, QIU L X, ZHANG D, et al. Aroma perception properties during oral processing of hot air-dried shrimp(Litopenaeus vannamei)[J]. Food and Fermentation Industries, 2024, 50(13): 223-231.

[52] GONZALEZ-ESTANOL K, PEDROTTI M, FONTOVA-CERD M, et al. Influence of chewing rate and food composition on in vivo aroma release and perception of composite foods[J]. Journal of Agricultural and Food Chemistry, 2024, 72(12): 6723-6734.

[53] LE QUÉRÉ J L, SCHOUMACKER R. Dynamic instrumental and sensory methods used to link aroma release and aroma perception: a review[J]. Molecules, 2023, 28(17): 6308.

[54] FOEGEDING E A, DRAKE M A. Invited review: sensory and mechanical properties of cheese texture[J]. Journal of Dairy Science, 2007, 90(4): 1611-1624.

[55] KING D A, HUNT M C, BARBUT S, et al. American meat science association guidelines for meat color mea-surement[J]. Meat and Muscle Biology, 2022, 6(4): 81-88.

[56] QI W L, JIANG Q Q, MA T Y, et al. Near-infrared spectroscopy combined with chemometrics for liquor product quality assessment: a review[J]. Foods, 2025, 14(17): 2992.

[57] LIU X H, HUANG M Z, TANG W Y, et al. Characteri-zation and exploration of the flavor profiles of green teas from different leaf maturity stages of Camellia sinensis cv. Fudingdabai using E-nose, E-tongue, and HS-GC-IMS combined with machine learning[J]. Foods, 2025, 14(16): 2861.

[58] CHEN C, LIU Z, YU H Y, et al. Flavoromic determination of lactones in cheddar cheese by GC-MS-olfactometry, aroma extract dilution analysis, aroma recombination and omission analysis[J]. Food Chemistry, 2022, 368: 130736.

[59] SUN J, CAI W, FENG T, et al. Revealing the flavor profile of citrus Pu-erh tea through GC-MS-O and untargeted metabolomics[J]. BEVERAGE PLANT RESEARCH, 2024, 4: e009.

[60] SHANG S Q, QING Y, ZHANG D P. Multi-omics technology-based flavor formation mechanisms and intelligent quality control research in strong-flavor Baijiu[J]. International Journal of Biology and Life Sciences, 2025, 11(1): 25-30.

[61] YANG Y, NIU C T, SHAN W X, et al. Physicochemical, flavor and microbial dynamic changes during low-salt doubanjiang (broad bean paste) fermentation[J]. Food Chemistry, 2021, 351: 128454.

[62] SWIEGERS J H, BARTOWSKY E J, HENSCHKE P A, et al. Yeast and bacterial modulation of wine aroma and flavour[J]. Australian Journal of Grape and Wine Research, 2005, 11(2): 139-173.

[63] STYGER G, PRIOR B, BAUER F F. Wine flavor and aroma[J]. Journal of Industrial Microbiology &Biotechnology, 2011, 38(9): 1145-1159.

[64] QIAN M, RUAN F X, ZHAO W H, et al. The dyna-mics of physicochemical properties, microbial community, and flavor metabolites during the fermentation of semi-dry Hakka rice wine and traditional sweet rice wine[J]. Food Chemistry, 2023, 416: 135844.

[65] HALKET J M, WATERMAN D, PRZYBOROWSKA A M, et al. Chemical derivatization and mass spectral libraries in metabolic profiling by GC/MS and LC/MS/MS[J]. Journal of Experimental Botany, 2005, 56(410): 219-243.

[66] CORDERO C, CAGLIERO C, LIBERTO E, et al. High concentration capacity sample preparation techniques to improve the informative potential of two-dimensional comprehensive gas chromatography-mass spectrometry: application to sensomics[J]. Journal of Chromatography A, 2013, 1318: 1-11.

[67] THOMAS-DANGUIN T, SINDING C, ROMAGNY S, et al. The perception of odor objects in everyday life: a review on the processing of odor mixtures[J]. Frontiers in Psychology, 2014, 5: 504.

[68] MANYI-LOH C E, NDIP R N, CLARKE A M. Volatile compounds in honey: a review on their involvement in aroma, botanical origin determination and potential biomedical activities[J]. International Journal of Molecular Sciences, 2011, 12(12): 9514-9532.

[69] VITA F, TAITI C, POMPEIANO A, et al. Aromatic and proteomic analyses corroborate the distinction between Mediterranean landraces and modern varieties of durum wheat[J]. Scientific Reports, 2016, 6: 34619.

[70] THOLL D, BOLAND W, HANSEL A, et al. Practical approaches to plant volatile analysis[J]. The Plant Journal, 2006, 45(4): 540-560.

[71] DASENAKI M E, THOMAIDIS N S. Quality and authenticity control of fruit juices: a review[J]. Molecules, 2019, 24(6): 1014.

[72] AMORIM K A, PASSOS L P, SILAS SOUZA A H, et al. Temporal dominance of Sensations: do different concepts of “dominance” affect the results?[J]. Food Research International, 2024, 194: 114902.

[73] LAZAROS K, KOUMADORAKIS D E, VLAMOS P, et al. Graph neural network approaches for single-cell data: a recent overview[J]. Neural Computing and Applications, 2024, 36(17): 9963-9987.

[74] ZHANG B H, WANG Y, WANG J P, et al. The establishment of ham grade, sensory scores and key flavor substances prediction models for Jinhua ham via E-nose combined with machine learning[J]. Food Chemistry, 2025, 472: 142847.

[75] 郭向阳, 宛晓春. 焙火程度对黄大茶挥发性香气成分的影响[J]. 现代食品科技, 2019, 35(10): 235-245.GUO X Y, WAN X C. Effect of roasting degree on volatile aroma components of large-leaf yellow tea[J]. Mo-dern Food Science and Technology, 2019, 35(10): 235-245.

[76] CAO N, YANG Y B, WANG L, et al. Characterization of the key flavor compounds contributing to the lingering cup aroma in sauce-flavor Baijiu using sensomics and flavoromics[J]. Journal of Food and Nutrition Research, 2024, 12(10): 461-470.

[77] GU Z P, XU Q Y, WANG X Y, et al. Food freshness prediction platform utilizing deep learning-based multimodal sensor fusion of volatile organic compounds and moisture distribution[J]. ACS Sensors, 2025, 10(4): 3091-3100.

[78] BORRS E, FERRÉ J, BOQUÉ R, et al. Olive oil sensory defects classification with data fusion of instrumental techniques and multivariate analysis (PLS-DA)[J]. Food Chemistry, 2016, 203: 314-322.

Application and Integration Prospects of Sensomics and Flavoromics in Food Flavor Research

WANG Bei* , DUAN Yunhan

(School of Food and Health, Beijing Technology and Business University, Beijing 100048, China)

Abstract： Flavor is a core element determining food quality. To solve the challenge of establishing a precise causal relationship between “chemical compositions” and “human perceptions” in complex food systems, sensomics and flavoromics have became two major paradigms in food flavor research in recent years. The core workflows, advantages, disadvantages, and future integration trends of these two research paradigms were systematically reviewed. Guided by human senses, sensomics established a causal chain between “key aroma-active compounds” and sensory attributes through gas chromatography-olfactometry (GC-O), precise quantification, and aroma recombination and omission experiments. It possessed strong mechanistic explanatory power but faced limitations such as low throughput, high quantification costs, and subjective evaluation. Flavoromics relied on high-throughput instruments to obtain untargeted “chemical fingerprints” and combined multivariate statistics to mine differential compounds. It had broad coverage and was commonly used for quality monitoring and traceability. However, its static analysis struggled to reflect the dynamic release in the oral cavity, and its correlation mining did not directly equate to causal verification. Given the complementarity between the “top-down causal verification” of sensomics and the “bottom-up correlation mining” of flavoromics, this study proposed that flavor science evolved toward an intelligent integration paradigm, which deeply merged the mechanistic verification logic of sensomics with the big data mining capabilities of flavoromics. It was suggested that future research relied on artificial intelligence and multi-omics technologies to construct standardized multimodal flavor databases and deepen the mechanistic analysis of spatiotemporal dynamic perception. This integrated paradigm was expected to complete the digital closed-loop between chemical compositions and sensory experiences, providing scientific support for the precise regulation and targeted creation of food flavors.

Keywords： food flavor; flavor compounds; sensomics; flavoromics; artificial intelligence

doi：10.12301/spxb202600113

文章编号：2095-6002(2026)03-0032-12

引用格式：王蓓,段云汉.感官组学与风味组学在食品风味研究中的应用与融合展望[J]. 食品科学技术学报,2026,44(3):32-43.

WANG Bei, DUAN Yunhan. Application and integration prospects of sensomics and flavoromics in food flavor research[J]. Journal of Food Science and Technology, 2026,44(3):32-43.

中图分类号： TS201.2

文献标志码： A

收稿日期： 2026-03-07

第一作者： 王 蓓,女,教授,博士,主要从事食品风味为主导的交叉学科方面的研究。

*通信作者。

(责任编辑：李 宁)