Figure 5a shows the results of this analysis, with the locations of the transcriptomic crests observed by Shen et al. (24) indicated. While the null results do not perfectly recapitulate the reported iPOP transcriptomic waves, it is notable that despite the lack of any true signal, two clear crests are apparent in our null study, one in the early 40s and one in the early 60s. Also notable is the correlation between the sample size in the sliding window and the number of changing molecules that are reported (Figure 5B); this phenomenon applies to both LOESS+DE-SWAN and DE-SWAN alone and is discussed further in Section 4. The appearance of two crests at these time points based on random noise calls into question the validity of the conclusion that they represent true biological trends.

There is a recent comment on pubpeer showing that the original Shen et al., paper replaces the real participant ages with fabricated ages during the LOESS step which is responsible for the "waves" https://pubpeer.com/publications/D2C91E5E2E55CE7452B12E44EBE844#2

Was this analysis done using the corrected code or the same code from the original paper. The authors may want to look into this and address it.