Modeling and Predicting the Activities of Trans-Acting Splicing Factors with Machine Learning

Abstract

Graphical Abstract Highlights d Most low-complexity domains in RBPs can regulate splicing when recruited to pre-mRNAs d Splicing regulatory activities of RBPs are mainly determined by sequence composition d Machine learning approach was developed to predict splicing regulatory activity of RBPs d The predictive model facilitates the design of artificial factors to manipulate splicing Correspondence wangzefeng@picb.an.cn In BriefAlternative splicing is mainly regulated by various trans-acting splicing factors that specifically bind cis-elements. A systematic survey was conducted to study splicing regulatory activities of many RBPs, providing a training set for a machine learning approach to predict splicing regulatory activities of endogenous RBPs and synthetic peptides. This study expanded the repertoire of potential splicing factors and revealed a direct link between the sequence composition and RBP activity. SUMMARYAlternative splicing (AS) is generally regulated by trans-splicing factors that specifically bind to cis-elements in pre-mRNAs. The human genome encodes $1,500 RNA binding proteins (RBPs) that potentially regulate AS, yet their functions remain largely unknown. To explore their potential activities, we fused the putative functional domains of RBPs to a sequence-specific RNA-binding domain and systemically analyzed how these engineered factors affect splicing. We discovered that $80% of lowcomplexity domains in endogenous RBPs displayed distinct context-dependent activities in regulating splicing, indicating that AS is under more extensive regulation than previously expected. We developed a machine learning approach to classify and predict the activities of RBPs based on their sequence compositions and further validated this model using endogenous RBPs and synthetic polypeptides. These results represent a systematic inspection, modeling, prediction, and validation of how RBP sequences affect their activities in controlling splicing, paving the way for de novo engineering of artificial splicing factors.

A widespread length-dependent splicing dysregulation in cancer

“…The gels were scanned with the ChemiDoc Touch Imaging System (Image Lab, Bio-Rad). The quantification of the bands was performed by the ImageJ software described previously ( 70 ).…”

Circular RNAs in eukaryotic cells: origin, characteristics, mechanisms of molecular functioning in human malignant diseases

“…Так как 7-метилгуанозин-модифицированные РНК экспортируются ограниченно в клетки, то это может объяснить роль длины РНК в контроле ядерного экспорта кольцевых РНК. Замкнутые малые ядерные РНК (snRNAs), так же как и длинные мРНК со шпильками, экспортируются через путь NXF1/NXT1, но если их длина сокращается (то есть мРНК без интронов), то они экспортируются через PHAX/CRM1-опосредованный путь [257]. Следовательно, локализация конечных кольцевых РНК зависит от их видов, образующихся в процессе бэксплайсинга (табл.…”

“…Впоследствии было доказано, что кольцевые РНК представляют собой одноцепочечные некодирующие РНК, ковалентно-замкнутые в непрерывную петлю (кольцо) [247], они участвуют в регуляции транскрипционной и посттранскрипционной экспрессии генов и эволюционно представлены в разных видах живых организмов от одноклеточных до млекопитающих [156,257,266]. Самая характерная особенность кольцевых РНК состоит в том, что они образуются в результате неканонического обратного сплайсинга или бэксплайсинга (англ.…”

Insights Into the Role of CircRNAs: Biogenesis, Characterization, Functional, and Clinical Impact in Human Malignancies

“…Circular RNAs are produced by non-canonical splicing events commonly known as backsplicing, which is considered an alternative splicing event. Although back splicing is regarded as an alternative splicing event, the molecular mechanism underlying the circular RNAs' biogenesis remains elusive (Mao et al, 2018). CircRNAs are derived from canonical splice sites and depend on canonical splicing machinery, which is usually inefficient to generate linear RNAs (Salzman et al, 2012;Jeck et al, 2013;Memczak et al, 2013;Ashwal-Fluss et al, 2014).…”