**2.4 Post-translational modification site prediction**

We used the MusiteDeep deep learning framework (https://github.com/duolinwa ng/MusiteDeep\_web) to search for the presence of possible post-translational modifications and identify how they affect the potential allergenicity of the study proteins [18]. The prediction models used are phosphorylation (Y, S, T); N-linked glycosylation (N); O-linked glycosylation (S, T); ubiquitination; N6-acetyllysine (K); Methylarginine (R); Methyllysine (K); Hydroxyproline (P) and Hydroxylysine (K) with a threshold value of 0.8.

S-nitrosylations and T-nitrations were also studied via the iSNO-AAPair tool (Y. Xu et al., 2013), which was used to predict cysteine S-nitrosylation sites (http://a pp.aporc.org/iSNO-AAPair) with a threshold value greater than 0.8. The GPS-YNO2 tool (Liu et al., 2011) was used to predict tyrosine nitration sites (http://yno2.biocuc koo.org).
