Figure 1 shows the histograms of the means of gene expressions from the autism group and the control group. One can see that the means of the two group have very similar distributions. Figure 2 shows the standard deviations of gene expression among the autism group and the control group, Again, the two groups seem to have similar standard deviations. To manage the gene expression magnitude, the gene expression levels were transformed by log2. Figure 3 shows the histogram of the z-scores of each of the genes, which indicates how different the means of the expression levels are between the two groups.
Figure 4 shows the histogram of the z-scores between the two batches for autism groups and the control groups, respectively. There are …show more content…
Using the Bonferroni correction, only two genes were detected to have significant differences in mean, as shown in Table 1, where p-value_adj_b is the adjusted p-value using Bonferroni correction. The Bonferroni correction can sometimes be very conservative, so an alternative, the Benjamini-Hochberg correction to control false discovery rates (FDR), was independently applied. Calculating the adjusted p-value≤0.05 resulted in 254 genes having a significant difference. Due to the number of the genes with differences, only the top 20 genes are listed in Table 2, where p-value_adj_bh is the adjusted p-value using Benjamini-Hochberg …show more content…
Depending on the multiple test correction methods, there are different number of genes expressed differently, as shown in Table 1 and Table 2. The Bonferroni correction can be very conservative since it assumes all tests are statistically independent, which may not be the case because certain genes have similar functionalities and/or a set of genes together controls some functions. Benjamini-Hochberg correction aims to consider the dependency of multiple tests and control the false detection rate. Given the false detection rate less than or equal 0.05, there are 254 genes showing significant expression differences when using Benjamini-Hochberg Correction, including the 2 gene expression differences detected when using Bonferroni correction with one of them being gene ZSCAN18. ZSCAN18 has been associated with affecting neurological development [5]. Nevertheless, the results support the assumption that autism has genetic base. Among the rest of the 253 differences, some of them, e.g., chromosome 15q11-13, have been identified from previous research [6]. The results here independently support the aforementioned