植物轉(zhuǎn)錄因子結(jié)合位點預(yù)測研究取得新突破

　　建模結(jié)果表明，DeepCNN在265個擬南芥數(shù)據(jù)集上都獲得了很高的預(yù)測精確度（平均AUC達(dá)0.96），闡明了其在植物TFBS預(yù)測方面的可行性。通過進(jìn)一步深入分析DeepCNN中卷積核的性質(zhì)，作者提供了模型的生物學(xué)可解釋性：DeepCNN不僅能學(xué)習(xí)到當(dāng)前轉(zhuǎn)錄因子在序列當(dāng)中的關(guān)鍵結(jié)合motif，而且能夠?qū)W習(xí)到與該轉(zhuǎn)錄因子共同協(xié)作的轉(zhuǎn)錄因子的結(jié)合motif。

 

　　最后當(dāng)使用遷移學(xué)習(xí)技術(shù)嘗試從計算的途徑解決目前植物TFBS研究問題的困難時，作者發(fā)現(xiàn)在不同的植物種類中，遷移學(xué)習(xí)的表現(xiàn)具有很大的不同。在水稻的十個TF中的三個都取得了比較好的預(yù)測效果，BZIP23 、ERF48和MADS29的 PPV（Positive predictive value）分別為0.752、0.951和0.816。而當(dāng)遷移到玉米和大豆中時，預(yù)測效果均不甚理想。這表明遷移學(xué)習(xí)在植物的跨物種轉(zhuǎn)錄因子結(jié)合位點預(yù)測問題上具有一定的可行性，但是未來我們?nèi)孕柙O(shè)計更加有效的遷移學(xué)習(xí)策略。

 

　　為了提供更方便、更優(yōu)質(zhì)的生物信息學(xué)服務(wù)，課題組為此具有高精確率辨別轉(zhuǎn)錄因子結(jié)合位點的深度卷積神經(jīng)網(wǎng)絡(luò)模型搭建了docker鏡像，通過下載該鏡像并在本地配置可以實現(xiàn)離線預(yù)測植物轉(zhuǎn)錄因子結(jié)合位點的預(yù)測功能（https://github.com/liulifenyf/TSPTFBS）。

 

　　【英文摘要】

 

　　Motivation: Both the lack or limitation of experimental data of transcription factor binding sites （TFBS） in plants and the independent evolutions of plant TFs make computational approaches for identifying plant TFBSs lagging behind the relevant human researches. Observing that TFs are highly conserved among plant species, here we first employ the deep convolutional neural network （DeepCNN） to build 265 Arabidopsis TFBS prediction models based on available DAP-seq （DNA affinity purification sequencing） datasets, and then transfer them into homologous TFs in other plants.

 

　　Results: DeepCNN not only achieves greater successes on Arabidopsis TFBS predictions when compared with gkm-SVM and MEME, but also has learned its known motif for most Arabidopsis TFs as well as cooperative TF motifs with PPI （protein-protein-interaction） evidences as its biological interpretability. Under the idea of transfer learning, trans-species prediction performances on ten TFs of other three plants of Oryza sativa, Zea mays and Glycine max demonstrate the feasibility of current strategy.

 

　　Availability and implementation: The trained 265 Arabidopsis TFBS prediction models were packaged in a Docker image named TSPTFBS, which is freely available on DockerHub at https://hub.docker.com/r/vanadiummm/tsptfbs. Source code and documentation are available on GitHub at: https://github.com/liulifenyf/TSPTFBS.

 

　　Contact: huxuehai@mail.hzau.edu.cn

 

　　原文鏈接：https://academic.oup.com/bioinformatics/article/37/2/260/6069568