Previous studies documented the important role of salience network, which consists of dorsal anterior cingulate cortex (dACC) and bilateral anterior insula (AI), in dynamic switching between activation of the default-mode network (DMN) and of the central executive network (CEN) to guide appropriate responses to salient stimuli¹. Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder associated with the altered functional connectivity of brain networks². Previous studies consistently reported the dysconnectivity of the DMN and salience network, and the distinctive between-network communications in ASD^3,4,5. However, studies examining the relationships between functional and structural connectivity (FC and SC) within and between networks are still lacking. This study aims to investigate the FC and SC of the salience network, DMN and CEN in ASD.We used resting-state fMRI (rs-fMRI) and diffusion spectrum imaging (DSI) to investigate the FC and SC, respectively. All data were acquired on a 3-Tesla Siemens Tim Trio system. Rs-fMRI (6-minute scan with eyes closed) data parameters: 180 echo planar imaging (EPI) volumes; repetition time (TR) = 2000 ms; echo time (TE) = 24 ms; flip angle = 90˚; field of view (FOV) = 256×256 mm^2; matrix size = 64×64; 34 axial slices acquired in an interleaved descending order; slice thickness = 3 mm; voxel size = 4×4×3 mm^3. Rs-fMRI data were preprocessed using the DPARSF toolbox with SPM8. Motion artifact was corrected by nuisance regression against 24-autoregressive motion parameters at an individual level. T1-weighted anatomical images were acquired using a 3D magnetization-prepared rapid gradient echo sequence, TR = 2530 ms; TE = 3.4 ms; slice thickness = 1.0 mm; matrix size = 256×256; and FOV = 256×256 mm^2. DSI data were acquired by a twice-refocused balanced echo diffusion echo planar imaging sequence, TR/TE = 9600/130 ms, imaging matrix size = 80 x 80, spatial resolution = 2.5 x 2.5 mm^2, and slice thickness = 2.5 mm. 102 diffusion encoding gradients with the maximum diffusion sensitivity bmax = 4000 s/mm^2 were applied to sample the grid points in a half sphere of the 3D q-space with |q| ≤ 3.6 units.After image preprocessing, the final sample included 78 with ASD and 69 typically developing controls (TD) with age, gender and FIQ matched (Table1). We conducted independent component analysis (ICA) using MELODIC in FSL 5.0.9, a data-driven approach to identify the three functional networks and the regions of interest (ROIs) (e.g. salience network: dACC, bilateral AI; DMN: medial prefrontal cortex (mPFC), bilateral precuneus; CEN: bilateral posterior parietal cortex and dorsal lateral prefrontal cortex) out of totally 55 components without making any prior assumptions (Figure1). For the SC analysis, the ROIs defined by ICA were further transformed to the DSI study-specific template made by the averaged individual DSI data to perform tractography and define tracts within the three networks (Figure2). The general fractional anisotropy values (GFA) of each tracts were sampled as the SC indices. For the FC analysis, 6-mm spheres centered on the peak coordinates of those ROIs were made, and the temporal correlations of blood-oxygen-level dependent signal between the ROIs within the three networks were calculated as the FC indices (Figure3). Between-group comparisons of the FC and SC were performaned using Multivariate Analysis of Covariance controlling for the age effects. Partial correlations were performed to investigate the relationships between FC and SC within the ASD and the TD group controlling for the age effects.The between-group comparisons showed that the GFA of the right DMN was significantly lower (p = .022), and the GFA of the left DMN was marginally significantly lower (p = .078) in the ASD group compared to the TD group. The comparisons of the within-group correlations between the FC and SC showed that the ASD group revealed overall more significant correlations between the FC and SC compared to TD, including the positive FC-SC correlations within the DMN. Notably, the FC of the right salience network are distinctively correlated to the SC of the DMN and CEN in the ASD group but not in the TD group (Figure4).In our study, we conducted a data-driven approach to investigate the FC and SC within three critical networks in ASD and TD. We found a significant weaker SC within the DMN, e.g. the cingulum bundles, in ASD compared to TD. Furthermore, we found significant positive correlations between the FC of the right salience network and the SC within the DMN and CEN. Given the role of salience network in modulating the switch between the DMN and CEN, our results might imply a distinctive FC-SC relationship across different networks in ASD.