SUPPLEMENTARY MATERIALS

Multisite Dataset

The multisite dataset used in this study was collected at the following sites included in this work: University of California, Irvine; University of California, Los Angeles and University of California, San Francisco; and Duke University, University of North Carolina, University of New Mexico, University of Iowa, and University of Minnesota. Table S1 shows the demographic information of the participant in each site. Overall, these sites have similar patients and healthy control distribution.

Signal-fluctuation-to-noise ratio (SFNR) techniques (Friedman et al., 2006) were used in the quality control steps to check reliability of the data across websites. SFNRs were calculated for the subjects’ EPI data sets as implemented using a “dataQuality” MATLAB package (http://cbi.nyu.edu/software/dataQuality.php), which is defined as the ratio of mean signal intensity across time and space to the average standard deviation of the same voxels’ time series in a region of interest in the center of the brain. Subjects who had an SFNR of less than 150 were excluded from the analysis (Damaraju et al., 2014). Later, the data went through a standard preprocessing pipeline and other quality control steps (subjects with a maximum root mean squared translation of more than 4 mm were also excluded).

Table S1: Demographic information on the subjects for each of the data sites.

Site / Scanner Type / # of subjects / # of Patients / Patients
M/F / Patients age mean (std) / #of Healthy / Healthy
M/F / Healthy age mean (std)
1. Duke / GE / 52 / 24 / 20/4 / 34.5(8.4) / 28 / 21/7 / 34.1(9.25)
2. Iowa / Siemens / 18 / 9 / 7/2 / 41.1(12.5) / 9 / 6/3 / 39.3(9.07)
3. UCI / Siemens / 54 / 26 / 21/5 / 44.1(11.9) / 28 / 22/6 / 43.1(12.7)
4.UCLA / Siemens / 51 / 23 / 18/5 / 34.9(11.9) / 28 / 23/5 / 35.6(11.6)
5.UCSF / Siemens / 27 / 13 / 11/4 / 36.7(10.7) / 14 / 11/3 / 36.4(9.1)
6.UMN / Siemens / 57 / 29 / 24/5 / 36.3(11.1) / 28 / 21/7 / 34.6(10.7)
7. UNM / Siemens / 55 / 27 / 20/7 / 38.6(11.7) / 28 / 24/4 / 36.8(10.4)

Results of Connectivity Strength

While comparing L_FNC and R_FNC among healthy controls, we found 623 out of 903 network pairs that are significantly (0.05 levels FDR corrected) connected in both FNCs, and 617 of them have the same sign (same directionality). By performing a paired t-test on L_FNC and R_FNC magnitudes and correcting with 0.05 levels FDR threshold, we found 145 network pairs exhibiting significant differences. Further, 102 of the pairs that were higher in R_FNC had stronger connectivity, while 43 were higher in L_FNC among the 617 previously selected network pairs. However, in the patients there were 556 network pairs that were significantly connected in both L_FNC and R_FNC, and 551 pairs had the same sign. Their paired t-tests revealed 74 network pairs with significant magnitude differences, of which 47 pairs were higher in R_FNC while 17 pairs were higher in L_FNC.

In comparing L_FNC and Inter_FNC, we analyzed LR (lower triangular) and RL (upper triangular) of the Inter_FNC separately. In the healthy controls and in lower triangular, 649 pairs out of 903 showed significant connectivity in both L_FNC and Inter_FNC (LR); of these, 647 had the same sign. Paired t-tests revealed that 40 regions were in favor of Inter_FNC, while 63 network-pairs were in favor of L_FNC. This number reduced to 573 significantly connected network-pairs, of which 572 had the same sign. Of these, 23 pairs favored Inter_FNC, and 41 of them favored L_FNC. In the healthy controls and in upper triangular, 637 pairs out of 903 showed significant connectivity in both L_FNC and Inter_FNC (LR), and 631 of those pairs had the same sign. Paired t-tests revealed 75 network pairs favoring Inter_FNC, while 58 had network pairs favoring L_FNC. This number reduced to 564 significantly connected pairs, of which 561 had the same sign, among which 28 pairs favor Inter_FNC and 31 of these favor L_FNC.

Comparison between R_FNC and Inter_FNC provided a similar pattern; in the healthy controls and in lower triangular part of Inter_FNC, 623 pairs out of 903 showed significant correlation in R_FNC and Inter_FNC (LR), of which 618 had the same sign. Paired t-tests revealed 34 network-pairs favoring Inter_FNC, while 119 pairs favored R_FNC. This number reduced to significantly to 562 correlated pairs, of which 557 had the same sign. Of these, 14 pairs favored inter-hemisphere, and 92 favored R_FNC. In healthy controls and in upper triangular part of Inter_FNC, 656 pairs out of 903 showed significant correlation in R_FNC and Inter_FNC (LR), of which 653 had the same sign. Paired t-test revealed 37 pairs favoring Inter_FNC, and 95 pairs favoring R_FNC. This number reduced to 567 significantly correlated network-pairs, of which 564 had the same sign. Of these, 17 pairs favored Inter_FNC and 63 of these favored L_FNC.

Table S2: Comparing Strength of different FNC types in patients and healthy control separately; at all FNCs motion and site ID were regressed out. For all pairs, we see a lack of asymmetries in patients when compared to healthy controls. Also, the right hemisphere interacts with itself more in comparison to the left hemisphere. Also, the right hemisphere has more strength in communication with itself in comparison to inter-hemispheric interaction.

Comparing Intra-Left and Intra-Right
L Patient / R Patient / L Healthy / R Healthy
FDR threshold one sample t-test / 0.0369 / 0.0347 / 0.0395 / 0.0393
# of Significant / 667 / 654 / 720 / 710
# of Significant in both / 556 / 623
# of Significant and same sign / 551 / 617
FDR threshold two sample t-test / 0.0053 / 0.0106
# of Intra Left<Intra Right significant / 47 / 102
# of Intra Left>Intra Right significant / 17 / 43
Comparing Intra-Left and Inter-Hemisphere (LR)
L Patient / LR Patient / L Healthy / LR Healthy
FDR threshold one sample t-test / 0.0369 / 0.0352 / 0.0395 / 0.0366
# of Significant / 667 / 637 / 720 / 710
# of Significant in both / 573 / 649
# of Significant and same sign / 572 / 647
FDR threshold two sample t-test / 0.0051 / 0.0078
# of Intra Left<Inter_LR significant / 23 / 40
# of Intra Left>Inter_LR significant / 41 / 63
Comparing Intra-Left and Inter-Hemisphere (RL)
L Patient / RL Patient / L Healthy / RL Healthy
FDR threshold one sample t-test / 0.0369 / 0.0344 / 0.0395 / 0.0390
# of Significant / 667 / 643 / 720 / 714
# of Significant in both / 564 / 637
# of Significant and same sign / 561 / 631
FDR threshold two sample t-test / 0.0053 / 0.0106
# of Intra Left<Inter_RL significant / 28 / 75
# of Intra Left>Inter_RL significant / 31 / 58
Comparing Intra-Right and Inter-Hemisphere (LR)
R Patient / LR Patient / R Healthy / LR Healthy
FDR threshold one sample t-test / 0.0347 / 0.0352 / 0.0393 / 0.0366
# of Significant / 654 / 637 / 710 / 710
# of Significant in both / 562 / 623
# of Significant and same sign / 557 / 618
FDR threshold two sample t-test / 0.0095 / 0.0122
# of Intra Right<Inter_LR significant / 14 / 34
# of Intra Right>Inter_LR significant / 92 / 119
Comparing Intra-Right and Inter-Hemisphere (RL)
R
Patient / RL Patient / R Healthy / RL Healthy
FDR threshold one sample t-test / 0.0347 / 0.0344 / 0.0393 / 0.0390
# of Significant / 654 / 643 / 710 / 714
# of Significant in both / 567 / 656
# of Significant and same sign / 564 / 653
FDR threshold two sample t-test / 0.0071 / 0.0096
# of Intra Right<Inter_RL significant / 17 / 37
# of Intra Right>Inter_RL significant / 63 / 95

Age and Gender Effects on Intra-Hemisphere and Inter-Hemisphere FNCs

Age effects on the three FNC types are presented in Figure S2, with increasing age mostly causing a decrease in correlations, particularly among visual networks and sensorimotor networks. Cognitive control networks show both increases and decreases in correlation with increasing age. Homotopic network correlation also shows decrease with increasing age in almost all networks; only one sensorimotor network exhibits increase in connectivity with aging, though the effect is not significant. This is consistent with Zuo et al. (2010), who found a decrease in homotopic connectivity in all regions, but sensorimotor in their voxel mirrored homotopic connectivity study.

We also checked the effect of gender, handedness, and cognitive scores on FNC values; our analysis indicated that gender is not a factor affecting any FNCs. Figure S3 presents the results. IC 74 seems to have the widest differences on the correlations with visual, auditory and sensorimotor networks. Handedness, cognitive scores, positive and negative symptoms scores do not have any significant effects (after 0.05 levels FDR correction) on any of the FNC types.

Regression analysis on intra-hemisphere and inter-hemisphere FNC types all reveal very similar patterns for age and gender. In all FNC types, a decrease in connectivity strength with increasing age is observed. This is consistent with findings in Allen et al. (2011), who used a larger dataset consisting only of health participants, and no major differences were observed between females and males.

Figure S1: Age effect on all FNC types, displayed as – log10(p-value)*sign(beta) format (at left column), after 0.05 levels FDR correction (at right column). General patterns look similar; network pairs have weaker connectivity with increasing age. Especially, visual sensorimotor networks diminished connectivity in elders; homotopic network pairs also show a decrease with increasing age.

Figure S2: Gender effects on all FNC types, displayed as – log10(p-value)*sign(beta) format (at left column), after 0.05 levels FDR correction (at right column). Gender is not a significant factor affecting FNCs; we have only a few regions with significant effects.

Effects of Cognitive Scores, Symptom Scores and Mean Antipsychotic Dose

We examined whether cognitive scores were significantly associated with intra- and inter- FNCs using a separate regression model with parameters cognitive scores, age, gender, diagnosis, handedness, motion and site ID. This analysis included 276 subjects for whom CMINDS scores were available, and it revealed no significant effect of cognitive scores on intra- or inter-FNCs or the left-right difference FNCs.

We also examined relationships between PANSS and intra- and inter-FNCs. We analyzed the effect of scores using a regression model including parameters PANSS (positive and negative scores in separate analysis), age, gender, handedness, motion and site ID using 148 subjects. Results did not indicate any significant relationship between PANNS and intra- and inter- FNCs or in left-right difference FNCs.

Finally, we examined if the mean antipsychotic dose has any effect on the intra- and inter-FNCs. Out of 151 subjects with schizophrenia, 129 of whom have the mean antipsychotic dose, were included in the analysis. We analyzed the effects of the mean antipsychotic dose using a regression model including CPZ values, age, gender, handedness, motion, and site ID as parameters. Results did not indicate any significant relationship between CPZ values and intra- and inter-FNCs or in left-right difference FNCs. The general pattern of the results did not change; therefore, we kept the analysis without CPZ values for simplicity.

Allen, E. A., Erhardt, E. B., Damaraju, E., Gruner, W., Segall, J. M., Silva, R. F., . . . Calhoun, V. D. (2011). A baseline for the multivariate comparison of resting-state networks. Front Syst Neurosci, 5, 2. doi: 10.3389/fnsys.2011.00002

Damaraju, E., Allen, E. A., Belger, A., Ford, J. M., McEwen, S., Mathalon, D. H., . . . Calhoun, V. D. (2014). Dynamic functional connectivity analysis reveals transient states of dysconnectivity in schizophrenia. Neuroimage-Clinical, 5, 298-308. doi: 10.1016/j.nicl.2014.07.003

Zuo, X. N., Kelly, C., Di Martino, A., Mennes, M., Margulies, D. S., Bangaru, S., . . . Milham, M. P. (2010). Growing together and growing apart: regional and sex differences in the lifespan developmental trajectories of functional homotopy. J Neurosci, 30(45), 15034-15043. doi: 10.1523/JNEUROSCI.2612-10.2010