The differentially expressed genes were found according to the m6A regulator expression values in normal and tumor samples. The 19 differentially expressed m6A-regulated genes with FDR < 0.05 and
\begin{document}$ { {\log}}_{ {2}}\left( {folding\;\; change}\right) < - {0.25} $\end{document} or
\begin{document}$ \geqslant {0.25} $\end{document} , including eight down-regulated m6A-regulated genes (FDR < 0.05 and
\begin{document}$ { {\log}}_{ {2}}\left( {folding\;\; change}\right) < - {0.25} $\end{document} ) and 11 up-regulated m6A-regulated genes (FDR < 0.05 and
\begin{document}$ { {\log}}_{ {2}}\left( {folding\;\;change}\right) \geqslant {0.25} $\end{document} ) were obtained. The remaining m6A-regulated genes were not differentially expressed in normal and tumor samples (supplementary Table S1). It can be seen from
Fig. 2A that the differentially expressed m6A-regulated genes
METTL3,
METTL5,
ZCCHC4,
YTHDF1,
HNRNPC,
LRPPRC,
HNRNPA2B1,
IGF2BP1,
IGF2BP3,
TRMT112 and
CPSF6 are highly expressed in tumor samples; genes
METTL14,
METTL16,
WTAP,
ZC3H13,
PCIF1,
SETD2,
FTO and
ALKBH3 were lowly expressed in tumor samples. The detailed information including average expression levels, differential fold, and corrected
P values for the 38 m6A-regulated genes were shown in the supplementary Table S1. Subsequently, we explored the expression relationship of 19 differentially expressed m6A regulators in lung adenocarcinoma. Notably, significant correlations were observed between the expression of methyltransferases, demethylases and binding proteins (
Fig. 2B). For example,
SETD2 was positively correlated with
METTL14, and the correlation reaches 0.69.
ZC3H13 was positively related to
METTL14,
HNRNPA2B1 and
SETD2, and the correlation coefficients were all greater than 0.5.
TRMT112 was negatively correlated with
METTL14, ZC3H13, CPSF6, SETD2 and
FTO, and
METTL5 was negatively correlated with
SETD2 and
FTO. Then, all tumor samples were classified into two classes based on 19 differentially expressed m6A regulators by consensus clustering.
Figs. 2C and
2D show that the relative change in area under the cumulative distribution function (CDF) curve is small when
k = 3, indicating that
k = 3 is a reasonable type. However,
Figs. 2E and
2F show that the sample size is too small for one of the three subtypes, so it is reasonable to consider splitting it into two subtypes. As in
Figs. 2G and 2H cannot well justify the division into two subtypes or three subtypes. To better illustrate the rationality of the typing, we used the Kaplan-Meier survival curve.
Figure 2I shows that the survival rate of Subtype B is better than that of Subtype A.
Figure 2J shows that when tumor samples were divided into three subtypes, the survival of Subtype C was intermediate between Subtypes A and B. Therefore, the division into two subtypes is reasonable.