面孔识别是人类高度发达的技能，在个体之间表现出较大的差异。以往的行为遗传学研究表明，面孔特异性识别能力的遗传度在0.6左右，但面孔识别能力的神经基础的遗传特征尚不清楚。功能磁共振成像研究已经发现了人类梭状回皮质区、尤其是右侧大脑半球的梭状回面孔区域，具有对面孔选择性的反应 (对面孔刺激的激活反应强于对其他非面孔刺激的反应)。因此，这个研究中，我们采用3种梭状回 (包括右侧FFA) 的多模态MRI数据——灰质体积、静息状低频振幅和任务态的面孔选择性反应——作为一个联合的表型来探索梭状回 (包含右侧梭状回面孔区在内)的特异性遗传模式；此外，我们还探讨了在大脑特异性高表达的基因对面孔网络特征的影响。 本文的主要结论是： 1. COMT多态性 (rs4633、rs4818和rs468)、NKAIN2 rs1842129多态性、KIAA0319 rs4504469多态性和NRCAM rs230052多态性的一组线性组合可能调节梭状回 (包含右侧梭状回面孔区) 的功能表型 (面孔选择性激活强度和静息态功能活动的相反方向的变化)。 2. CNTAPN2多态性 (rs253897和rs10246245)、PICALM rs3551179多态性、TCF4 rs2558182多态性、KIAA0319 rs450464多态性和TMM40 rs157580多态性的一组线性组合可能调节梭状回 (包含右侧梭状回面孔区) 的另一种形式的功能表型 (面孔选择性激活强度和静息态功能活动的相同方向的变化)。 3. APOE rs7412多态性、NKAIN2 rs1842129多态性和NRCAM rs2300005多态性的线性组合可能调节梭状回 (包含右侧面孔选择区) 的结构表型 (灰质体积)。 4. 客体或面孔选择区的结构和功能MRI表型的遗传调控模式可能是分离的。即，调控面孔选择区结构表型的遗传成分对功能表型的贡献小，而调控面孔选择区功能表型的遗传成分对结构表型的贡献小。 5. 我们在一组探索性数据集和两组验证性数据集中都分别发现了一组和环核苷酸信号转导分子环路有关的基因分数和面孔网络的最短路径成正相关，而和聚合系数和度呈现负相关。 综上，我们采用了一种增强统计效力的多元分析方法，从分子生物学的角度探讨和验证了客体分类的脑区和面孔特异性脑区的功能和结构的遗传调控模式、发现了可能对面孔识别选择性脑区具有遗传调控作用的基因组合，这同时也对我们探讨其他功能特异性脑区的遗传调控模式提供了一个窗口，对我们未来的研究中探索相关的神经精神疾病的可能病因也提供了有价值的参考。

Face recognition is a highly developed skill in humans that shows considerable variability between individuals. Previous behavioral genetics studies have shown that face-specific recognition ability is heritable, and functional magnetic resonance imaging (fMRI) studies have identified a cortical region in human fusiform gyrus especially in the right hemisphere (fusiform face area, rFFA) that responds selectively to faces. Here, we asked whether the fusiform (including the rFFA) is heritable by using multimodal MRI data of the fusiform/rFFA (multimodal-MRI data), including gray matter volume (GM), fractional amplitude of low-frequency fluctuations from resting-state fMRI (fALFF), and face-selective responses from task-state fMRI, as a combined phenotype to explore specific genetic variants contributing to the fusiform or the rFFA. We also explored the effects of brain-specific genes on the face network characteristics. First, we selected 26 single nucleotide polymorphisms (SNPs) previously shown to be related to visual cognition as candidate genetic variants. Then, to identify SNPs that might modulate the fusiform (including the rFFA) we preformed a partial least squares correspondence analysis (PLSCA) between SNP data and the multimodal-fusiform-MRI data from 338 Han Chinese adults (mean age = 20.45 ± 0.96 years, 135 males). We found three significant components of SNP-fusiform associations. The first partial least squares component (PLS1) represented a profile (linear combination) of COMT polymorphisms (rs4633, rs4818, and rs4680), a NKAIN2 rs1842129 polymorphism, a NRCAM rs2300052 polymorphisms, and a KIA0319 rs4504469 polymorphism that were positively correlated with FS, negatively correlated with fALFF, and uncorrelated with GMV of the fusiform. The PLS2 represented a profile of CNTNAP2 polymorphisms (rs2538976 and rs10246245), a PICALM rs3851179 polymorphism, a TCF4 rs2958182 polymorphism, a KIA0319 rs4504469 polymorphism and a TOMM40 rs157580 polymorphism that were positively correlated with both the FS and the fALFF, but uncorrelated with GMV of the fusiform. PLS3 identified a profile of APOE rs7412 polymorphism, a NKAIN2 rs1842129 polymorphism and a NRCAM rs230005 polymorphism that were uncorrelated with the FS and the fALFF, but positively correlated with GMV of the fusiform. Next, to validate our results, we conducted another PLS analysis between expression values of candidate genes from Allen Gene Expression Dataset at 76 sample sites of the bilateral fusiform and the multimodal-rFFA-MRI data from the 76 brain regions of interest (ROIs) in the fusiform. We confirmed the effects of the APOE and NKAIN2 on GMV of the fusiform, but only partially validated the effects of genetic modulation on the functional fusiform. Further, to examine which specific single nucleotide polymorphisms (SNPs) contribute to the face selective region of rFFA, we sequenced the participants’ candidate SNPs selected from the SNP-fusiform association analyses and then performed PLSCA between the SNP data and the multimodal-rFFA-MRI data. We found that the functional rFFA was modulated by gene COMT and gene NKNIA2 via rs6269-rs4633-rs4818-rs4680 haplotypes and rs1842129 polymorphism, and the structural rFFA was modulated by gene APOE, NKAIN2 and NRCAM. Finally, we conducted PLS between brain-specific genes and face network parameters (average path length, clustering coefficients, betweenness centrality and degree). We identified a significant gene component (related to cyclic-nucleotide-mediated signaling) that positively associated with average path lengths and negatively associated with clustering coefficients, betweenness centrality and degree. In sum, our study provides the first empirical evidence on the genetic origin of the object-recognition region, especially the face-specific recognition region of rFFA, which sheds light on how the functional specificity of the rFFA is formed and illustrates a new way of investigating the origin of other function-specialized brain regions.