|
| Status |
Public on May 05, 2021 |
| Title |
Convolutional neural network modelling: advancing identification of true mRNA cleavage sites |
| Organisms |
Solanum tuberosum; Phytophthora infestans |
| Experiment type |
Expression profiling by high throughput sequencing
Non-coding RNA profiling by high throughput sequencing
|
| Summary |
Degradome sequencing is commonly used to generate high-throughput information on mRNA cleavages by small RNAs. Here we developed an extension module based on a deep learning convolutional neural network (CNN) in a machine learning environment to discriminate false from true cleavage sites applied on datasets from potato (Solanum tuberosum, St) and the oomycete pathogen Phytophthora infestans (Pi). The core of the CNN module is a stochastic gradient descent optimizer with cyclical learning rate (CLR) which together with Bayesian optimization scored a validation accuracy of 100%. To verify the recognition of cleavage sites we applied the module on Arabidopsis thaliana microRNA cleavages. Our module managed to recognize all cleavages, confirming the reliability of the module. When applying this new model to evaluate our data, 7.3% of all cleavage windows represented true cleavages distributed as 214 sites in P. infestans and 444 sites in potato. The sRNA landscape of the potato-P. infestans interaction is complex with uneven sRNA production and cleavage regions. In total, 222 endogenous Pi-sRNAs, 565 endogenous St-sRNAs, 91 trans-acting Pi-sRNAs and 14 trans-acting St-sRNAs were discovered from our datasets. Groups of self-regulatory sRNAs, and sRNAs generated from effector sequences like RXLR and Crinklers suggest dual effector functions. In the potato genome, resistance genes were most targeted mainly as self-regulation but also as results of a trans-action event. Our new analytic model is freely accessible for anyone working on complex biological systems.
|
| |
|
| Overall design |
9 small RNA samples sequenced from Co-IP. 3 replicates for each of the following: StAgo1-GFP H2O, StAgo1-GFP 11388, StAgo1-GFP 88069. 4 small RNA libraries were sequenced. 2 replicates from each of the following: Potato leaves (cv. Sarpo Mira) inoculated with H20 or 88069. 10 degradome libraries were sequenced. 2 replicates from each of the following: Mycelia of P. infestans strains 88069 or PiAgo1-GFP. Leaf samples of Potato (cv. Bintje) inoculated with H2O. Infected leaf samples with P. infestans strains 88069 or PiAgo1-GFP.
|
| |
|
| Contributor(s) |
Hodén KP, Hu X, Martinez-Arias G, Dixelius C |
| Citation(s) |
33924042 |
| Submission date |
Dec 16, 2020 |
| Last update date |
May 05, 2021 |
| Contact name |
Kristian Erik Persson Hodén |
| E-mail(s) |
kristian.hoden@gmail.com
|
| Organization name |
Swedish University of Agricultural Sciences
|
| Department |
Plant Biology
|
| Lab |
Christina Dixelius
|
| Street address |
Almas allé 5
|
| City |
Uppsala |
| ZIP/Postal code |
75651 |
| Country |
Sweden |
| |
|
| Platforms (4)
|
| GPL22257 |
Illumina HiSeq 2500 (Solanum tuberosum) |
| GPL25509 |
Ion Torrent Proton (Phytophthora infestans; Solanum tuberosum) |
| GPL29221 |
Illumina HiSeq 2500 (Phytophthora infestans; Solanum tuberosum) |
| GPL29506 |
Ion Torrent Proton (Solanum tuberosum) |
|
| Samples (35)
|
|
| Relations |
| BioProject |
PRJNA685865 |
| SRA |
SRP298275 |
Less...
More...