A high prediction accuracy was obtained, revealing a close association between histone modifications and exon expression. Consistently, based on quantitative RT-PCR and ChIP, Ref found that the distributions of a set of histone modifications on alterative spliced exons, including H3K36me3, H3K4me1, H3K27me3, H3K4me3 and H3K9me1, significantly changed accordingly with the differential exon expression in different cell types. These biological experiments further verified our founding. Recently, a rule-based model was trained to predict exon inclusion levels from the histone modification combinations. To further explore the relationship among histone modifications, gene expression and exon expression, we constructed an interaction network based on partial correlations. The network indicated that aside from gene expression, two histone modifications contribute directly to exon expression. This suggests that such two histone modifications could account for exon inclusion with transcription effects regressed away. Consistently, Luco et al. utilized biological experiments to validate the causal role of histone modifications in alternative splice site selections, and indicated that H3K36me3 could interact with PTB protein to regulate alternative splicing. A preliminary study by Han et al. utilized a Bayesian network to infer the histone modification regulatory network on gene promoters, while this work investigates the histone modification interaction relationships on exons. A recent study indicated that the cross-talk between histone modifications could evolve over different genomic elements. Consistently, Zippo et al. found that if a histone phosphorylation H3S10ph respectively happens on the enhancer and promoter, it will induce different histone modifications, and consequently, result in different gene expression status. Compared to the network on promoters, different histone modification interactions were actually revealed on exons. This possibly results from three mechanisms. Firstly, promoters and exons could recruit different protein complexes. Wang et al. revealed distinct distributions of histone acetyltransferase and deacetyltransferase on promoters and transcribed regions, including p300, CBP, PCAF, MOF, Tip60, HDAC1, HDAC2, HDAC3 and HDAC6. These distinct distributions could contribute to the different histone modification interaction relationships on promoters and exons. Secondly, the specific sequence KRX-0401 clinical trial features on two different genomic elements could result in the different regulatory relationships. For example, compared to promoters, exons have specific sequence Exon Splicing Silencer. Luco et al. indicated that ESS could recruit PTB; furthermore, PTB interacts with MRG15, a component of the retinoblastoma binding protein 2. Thirdly, genomic elements could exert different chromatin structure, which associate with different chromatin remodeling proteins. Some chromatin remodelers have been indicated to function as or associate with histone modifiers. In the partial correlation interaction network, specific combinations of histone modifications were identified, some contribute directly to exon expression and others regulate the differentiation of histone modification levels. Traditionally, proponents of the histone code hypothesis have held that specific combinations of histone modifications specify distinct downstream effects in a BEZ235 synergistic fashion, while other studies have only revealed the simple additive consequence of histone modifications. Consistently with the latter hypothesis, our results suggested that combinations of histone modifications on exons regulate exon expression cumulatively rather than synergistically. In addition, the existence of redundancy was observed in the histone modifications on exon regions. Our results suggested that only two histone modifications could faithfully predict exon expression levels.