A deep learning framework to predict binding preference of RNA constituents on protein surface

Jordy Homing Lam; Yu Li; Lizhe Zhu; Ramzan Umarov; Hanlun Jiang; Amélie Héliou; Fu Kit Sheong; Tianyun Liu; Yongkang Long; Yunfei Li; Liang Fang; Russ B. Altman; Wei Chen; Xuhui Huang; Xin Gao

doi:10.1038/s41467-019-12920-0

A deep learning framework to predict binding preference of RNA constituents on protein surface

Jordy Homing Lam, Yu Li, Lizhe Zhu, Ramzan Umarov, Hanlun Jiang, Amélie Héliou, Fu Kit Sheong, Tianyun Liu, Yongkang Long, Yunfei Li, Liang Fang, Russ B. Altman, Wei Chen, Xuhui Huang, Xin Gao

Research output: Contribution to journal › Article › peer-review

75 Scopus citations

Abstract

Protein-RNA interaction plays important roles in post-transcriptional regulation. However, the task of predicting these interactions given a protein structure is difficult. Here we show that, by leveraging a deep learning model NucleicNet, attributes such as binding preference of RNA backbone constituents and different bases can be predicted from local physicochemical characteristics of protein structure surface. On a diverse set of challenging RNA-binding proteins, including Fem-3-binding-factor 2, Argonaute 2 and Ribonuclease III, NucleicNet can accurately recover interaction modes discovered by structural biology experiments. Furthermore, we show that, without seeing any in vitro or in vivo assay data, NucleicNet can still achieve consistency with experiments, including RNAcompete, Immunoprecipitation Assay, and siRNA Knockdown Benchmark. NucleicNet can thus serve to provide quantitative fitness of RNA sequences for given binding pockets or to predict potential binding pockets and binding RNAs for previously unknown RNA binding proteins.

Original language	English (US)
Journal	Nature Communications
Volume	10
Issue number	1
DOIs	https://doi.org/10.1038/s41467-019-12920-0
State	Published - Oct 30 2019

Bibliographical note

KAUST Repository Item: Exported on 2020-10-01
Acknowledged KAUST grant number(s): FCC/1/1976-18-01, FCC/1/1976-23-01, FCC/1/1976-25-01, FCC/1/1976-26-01, URF/1/3007-01
Acknowledgements: We are grateful to Wei Wang for helpful discussions. Figure 1 was created by Heno Hwang, scientific illustrator at King Abdullah University of Science and Technology (KAUST). This work was supported by grants from KAUST to X.G. (BAS/1/1624-01, FCC/1/1976-18-01, FCC/1/1976-23-01, FCC/1/1976-25-01, FCC/1/1976-26-01, and FCS/1/4102-02-01) and funding from the KAUST to X.G. and X.H. (URF/1/3007-01). The Hong Kong Research Grant Council (HKUST C6009-15G, AoE/M-09/12, and AoE/P-705/16) and Innovation and Technology Commission (ITCPD/17-9 and ITC-CNERC14SC01) to X.H.; L.F., Y.F.L., and W.C. were supported by Research Grant from Science and Technology Innovation Commission of Shenzhen Municipal Government (No. KQTD20180411143432337 and JCYJ20170307105752508). Part of bioinformatics analysis was supported by the Center for Computational Science and Engineering of Southern University of Science and Technology.

Access to Document

10.1038/s41467-019-12920-0

https://repository.kaust.edu.sa/bitstream/10754/659549/1/s41467-019-12920-0.pdf

Cite this

@article{0a03efcf06bd4fdeb780ffa7950f4440,

title = "A deep learning framework to predict binding preference of RNA constituents on protein surface",

abstract = "Protein-RNA interaction plays important roles in post-transcriptional regulation. However, the task of predicting these interactions given a protein structure is difficult. Here we show that, by leveraging a deep learning model NucleicNet, attributes such as binding preference of RNA backbone constituents and different bases can be predicted from local physicochemical characteristics of protein structure surface. On a diverse set of challenging RNA-binding proteins, including Fem-3-binding-factor 2, Argonaute 2 and Ribonuclease III, NucleicNet can accurately recover interaction modes discovered by structural biology experiments. Furthermore, we show that, without seeing any in vitro or in vivo assay data, NucleicNet can still achieve consistency with experiments, including RNAcompete, Immunoprecipitation Assay, and siRNA Knockdown Benchmark. NucleicNet can thus serve to provide quantitative fitness of RNA sequences for given binding pockets or to predict potential binding pockets and binding RNAs for previously unknown RNA binding proteins.",

author = "Lam, {Jordy Homing} and Yu Li and Lizhe Zhu and Ramzan Umarov and Hanlun Jiang and Am{\'e}lie H{\'e}liou and Sheong, {Fu Kit} and Tianyun Liu and Yongkang Long and Yunfei Li and Liang Fang and Altman, {Russ B.} and Wei Chen and Xuhui Huang and Xin Gao",

note = "KAUST Repository Item: Exported on 2020-10-01 Acknowledged KAUST grant number(s): FCC/1/1976-18-01, FCC/1/1976-23-01, FCC/1/1976-25-01, FCC/1/1976-26-01, URF/1/3007-01 Acknowledgements: We are grateful to Wei Wang for helpful discussions. Figure 1 was created by Heno Hwang, scientific illustrator at King Abdullah University of Science and Technology (KAUST). This work was supported by grants from KAUST to X.G. (BAS/1/1624-01, FCC/1/1976-18-01, FCC/1/1976-23-01, FCC/1/1976-25-01, FCC/1/1976-26-01, and FCS/1/4102-02-01) and funding from the KAUST to X.G. and X.H. (URF/1/3007-01). The Hong Kong Research Grant Council (HKUST C6009-15G, AoE/M-09/12, and AoE/P-705/16) and Innovation and Technology Commission (ITCPD/17-9 and ITC-CNERC14SC01) to X.H.; L.F., Y.F.L., and W.C. were supported by Research Grant from Science and Technology Innovation Commission of Shenzhen Municipal Government (No. KQTD20180411143432337 and JCYJ20170307105752508). Part of bioinformatics analysis was supported by the Center for Computational Science and Engineering of Southern University of Science and Technology.",

year = "2019",

month = oct,

day = "30",

doi = "10.1038/s41467-019-12920-0",

language = "English (US)",

volume = "10",

journal = "Nature Communications",

issn = "2041-1723",

publisher = "Nature Research",

number = "1",

}

TY - JOUR

T1 - A deep learning framework to predict binding preference of RNA constituents on protein surface

AU - Lam, Jordy Homing

AU - Li, Yu

AU - Zhu, Lizhe

AU - Umarov, Ramzan

AU - Jiang, Hanlun

AU - Héliou, Amélie

AU - Sheong, Fu Kit

AU - Liu, Tianyun

AU - Long, Yongkang

AU - Li, Yunfei

AU - Fang, Liang

AU - Altman, Russ B.

AU - Chen, Wei

AU - Huang, Xuhui

AU - Gao, Xin

N1 - KAUST Repository Item: Exported on 2020-10-01 Acknowledged KAUST grant number(s): FCC/1/1976-18-01, FCC/1/1976-23-01, FCC/1/1976-25-01, FCC/1/1976-26-01, URF/1/3007-01 Acknowledgements: We are grateful to Wei Wang for helpful discussions. Figure 1 was created by Heno Hwang, scientific illustrator at King Abdullah University of Science and Technology (KAUST). This work was supported by grants from KAUST to X.G. (BAS/1/1624-01, FCC/1/1976-18-01, FCC/1/1976-23-01, FCC/1/1976-25-01, FCC/1/1976-26-01, and FCS/1/4102-02-01) and funding from the KAUST to X.G. and X.H. (URF/1/3007-01). The Hong Kong Research Grant Council (HKUST C6009-15G, AoE/M-09/12, and AoE/P-705/16) and Innovation and Technology Commission (ITCPD/17-9 and ITC-CNERC14SC01) to X.H.; L.F., Y.F.L., and W.C. were supported by Research Grant from Science and Technology Innovation Commission of Shenzhen Municipal Government (No. KQTD20180411143432337 and JCYJ20170307105752508). Part of bioinformatics analysis was supported by the Center for Computational Science and Engineering of Southern University of Science and Technology.

PY - 2019/10/30

Y1 - 2019/10/30

N2 - Protein-RNA interaction plays important roles in post-transcriptional regulation. However, the task of predicting these interactions given a protein structure is difficult. Here we show that, by leveraging a deep learning model NucleicNet, attributes such as binding preference of RNA backbone constituents and different bases can be predicted from local physicochemical characteristics of protein structure surface. On a diverse set of challenging RNA-binding proteins, including Fem-3-binding-factor 2, Argonaute 2 and Ribonuclease III, NucleicNet can accurately recover interaction modes discovered by structural biology experiments. Furthermore, we show that, without seeing any in vitro or in vivo assay data, NucleicNet can still achieve consistency with experiments, including RNAcompete, Immunoprecipitation Assay, and siRNA Knockdown Benchmark. NucleicNet can thus serve to provide quantitative fitness of RNA sequences for given binding pockets or to predict potential binding pockets and binding RNAs for previously unknown RNA binding proteins.

AB - Protein-RNA interaction plays important roles in post-transcriptional regulation. However, the task of predicting these interactions given a protein structure is difficult. Here we show that, by leveraging a deep learning model NucleicNet, attributes such as binding preference of RNA backbone constituents and different bases can be predicted from local physicochemical characteristics of protein structure surface. On a diverse set of challenging RNA-binding proteins, including Fem-3-binding-factor 2, Argonaute 2 and Ribonuclease III, NucleicNet can accurately recover interaction modes discovered by structural biology experiments. Furthermore, we show that, without seeing any in vitro or in vivo assay data, NucleicNet can still achieve consistency with experiments, including RNAcompete, Immunoprecipitation Assay, and siRNA Knockdown Benchmark. NucleicNet can thus serve to provide quantitative fitness of RNA sequences for given binding pockets or to predict potential binding pockets and binding RNAs for previously unknown RNA binding proteins.

UR - http://hdl.handle.net/10754/659549

UR - http://www.nature.com/articles/s41467-019-12920-0

UR - http://www.scopus.com/inward/record.url?scp=85074261461&partnerID=8YFLogxK

U2 - 10.1038/s41467-019-12920-0

DO - 10.1038/s41467-019-12920-0

M3 - Article

C2 - 31666519

SN - 2041-1723

VL - 10

JO - Nature Communications

JF - Nature Communications

IS - 1

ER -

A deep learning framework to predict binding preference of RNA constituents on protein surface

Abstract

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this