 Research
 Open Access
 Published:
LPIdeepGBDT: a multiplelayer deep framework based on gradient boosting decision trees for lncRNA–protein interaction identification
BMC Bioinformatics volume 22, Article number: 479 (2021)
Abstract
Background
Long noncoding RNAs (lncRNAs) play important roles in various biological and pathological processes. Discovery of lncRNA–protein interactions (LPIs) contributes to understand the biological functions and mechanisms of lncRNAs. Although wet experiments find a few interactions between lncRNAs and proteins, experimental techniques are costly and timeconsuming. Therefore, computational methods are increasingly exploited to uncover the possible associations. However, existing computational methods have several limitations. First, majority of them were measured based on one simple dataset, which may result in the prediction bias. Second, few of them are applied to identify relevant data for new lncRNAs (or proteins). Finally, they failed to utilize diverse biological information of lncRNAs and proteins.
Results
Under the feedforward deep architecture based on gradient boosting decision trees (LPIdeepGBDT), this work focuses on classify unobserved LPIs. First, three human LPI datasets and two plant LPI datasets are arranged. Second, the biological features of lncRNAs and proteins are extracted by Pyfeat and BioProt, respectively. Thirdly, the features are dimensionally reduced and concatenated as a vector to represent an lncRNA–protein pair. Finally, a deep architecture composed of forward mappings and inverse mappings is developed to predict underlying linkages between lncRNAs and proteins. LPIdeepGBDT is compared with five classical LPI prediction models (LPIBLS, LPICatBoost, PLIPCOM, LPISKF, and LPIHNM) under three cross validations on lncRNAs, proteins, lncRNA–protein pairs, respectively. It obtains the best average AUC and AUPR values under the majority of situations, significantly outperforming other five LPI identification methods. That is, AUCs computed by LPIdeepGBDT are 0.8321, 0.6815, and 0.9073, respectively and AUPRs are 0.8095, 0.6771, and 0.8849, respectively. The results demonstrate the powerful classification ability of LPIdeepGBDT. Case study analyses show that there may be interactions between GAS5 and Q15717, RAB30AS1 and O00425, and LINC01572 and P35637.
Conclusions
Integrating ensemble learning and hierarchical distributed representations and building a multiplelayered deep architecture, this work improves LPI prediction performance as well as effectively probes interaction data for new lncRNAs/proteins.
Introduction
Long noncoding RNAs (lncRNAs) are a class of important noncoding RNAs with the length more than 200 nucleotides. The class of RNAs have been reported to have dense associations with multiple biological processes including RNA splicing, transcriptional regulation, and cell cycle [1,2,3]. More importantly, the mutations and dysregulations of lncRNAs have important affects on multiple cancers [4, 5], for instance, lung cancer [6], colon cancer [7], and prostate cancer [8]. For example, lncRNAs UCA1, PCA3, and HOTAIR have been used as possible biomarkers of bladder cancer detection, prostate cancer aggressiveness, and hepatocellular carcinoma recurrence, respectively [9,10,11]. Although lncRNAs have been intensively investigated, functions and molecular mechanisms of lncRNAs still largely remain elusive [2, 12]. Recent researches have revealed that lncRNAs densely link to the corresponding bindingproteins. Therefore, the identification of the binding proteins for lncRNAs is urgent for better understanding the biological functions and molecular mechanisms of lncRNAs [1].
Although wet experiments for lncRNA–protein Interaction (LPI) discovery have been designed, computational methods are appealing to infer the relevances between lncRNAs and proteins [13]. The computational methods can be roughly divided into two categories: networkbased methods and machine learningbased methods. Networkbased LPI inference methods integrated various biological data and designed network propagation methods to find potential LPIs in the heterogeneous lncRNA–protein network. For example, Li et al. [14] proposed a random walk with restartbased LPI prediction model. Zhou et al. [15] took miRNAs as mediators to predict LPIs in a heterogeneous network (LPIHNM). Yang et al. [16] used the HeteSim algorithm to compute the associated scores between lncRNAs and proteins. Zhao et al. [17], Ge et al. [18], and Xie et al. [19] explored a few bipartite network projectionbased recommendation techniques to compute the interaction probabilities between lncRNAs and proteins. Zhang et al. [20] explored a novel LPI prediction framework combining a linear neighborhood propagation algorithm. Zhou et al. [21] combined similarity kernel fusion and Laplacian regularized least squares to find unobserved LPIs (LPISKF).
Machine learningbased LPI inference methods characterized the biological features of lncRNAs and proteins and exploited machine learning algorithms to probe LPI candidates [22]. Machine learningbased LPI prediction methods contain matrix factorization techniques and ensemble learning techniques [23]. Matrix factorizationbased LPI prediction approaches used various matrix factorization techniques. Liu et al. [24] identified new LPIs combing neighborhood regularized logistic matrix factorization. Zhao et al. [25] inferred LPI candidates combining the neighborhood regularized logistic matrix factorization model and random walk. Zhang et al. [26] proposed a graph regularized nonnegative matrix factorization method to uncover unobserved LPIs.
Ensemble learningbased LPI inference methods utilized diverse ensemble techniques. Zhang et al. [27] exploited an ensemble learning model to discover the interactions between lncRNAs and proteins. Liu et al. [24] designed three ensemble strategies to predict LPIs based on support vector machine, random forest and extreme gradient boosting, respectively. Deng et al. [1] extracted HeteSim features and diffusion features of lncRNAs and proteins and constructed a gradient tree boostingbased LPI prediction algorithm (PLIPCOM). Fan and Zhang [28] explored a stacked ensemblebased LPI classification model via logistical regression (LPIBLS). Deng et al. [29] proposed a gradient boosted regression tree for finding possible LPIs. Wekesa et al. [30] designed a categorical boostingbased LPI discovery framework (LPICatBoost). In addition, deep learning (such as deep graph neural network [31]) is increasingly developed to identify LPI candidates.
Computational methods effectively identified potential LPIs. However, there are a few problems to solve. First, the majority of computational models were evaluated on one dataset, which may result in predictive bias. Second, they were not used to infer potential proteins (or lncRNAs) associated with a new lncRNA (or protein). Finally, their prediction performance need to further improve.
To solve the above problems, in this study, inspired by Gradient Boosting Decision Trees (GBDT) provided by Feng et al. [32], we exploit a multiplelayer Deep structure with GBDT to predict unobserved LPIs (LPIdeepGBDT). First, five LPI datasets are constructed. Second, lncRNA and protein features are extracted by Pyfeat and BioProt, respectively. Third, a feature vector is built to represent an lncRNA–protein pair. Finally, a multiplelayer deep architecture integrating tree ensembles and hierarchical distributed representations is developed to classify lncRNA–protein pairs.
The remaining of this manuscript is organized as follows. “Materials and methods” section describes data resources and the LPIdeepGBDT framework. “Results” section illustrates the results from a series of experiments. “Discussion and further research” section discusses the LPIdeepGBDT method and provides directions for further research.
Materials and methods
Data preparation
In this manuscript, we collect three human LPI datasets and two plant LPI datasets. Dataset 1 provided by Li et al [14] contains 3,487 LPIs from 938 lncRNAs and 59 proteins. 3,479 LPIs between 935 lncRNAs and 59 proteins are finally obtained by removing the lncRNAs without sequence information in the NONCODE [33], NPInter [34] and UniProt [35] databases.
Dataset 2 build by Zheng et al. [36] contains human 4,467 LPIs between 1,050 lncRNAs and 84 proteins. 3,265 LPIs from 885 lncRNAs and 84 proteins are extracted after removing the lncRNAs without any sequence information. Dataset 3 constructed by Zhang et al. [20] contains 4,158 LPIs between 990 lncRNAs and 27 proteins.
Datasets 4 provides 948 Arabidopsis thaliana LPIs from 109 lncRNAs and 35 proteins. Dataset 5 provides 22,133 Zea mays LPIs from 1,704 lncRNAs and 42 proteins. The sequence information of two entities is downloaded from the PlncRNADB database [37] and LPIs are extracted at http://bis.zju.edu.cn/PlncRNADB/. The details are described in Table 1.
We denote an LPI network via a matrix Y:
Overview of LPIdeepGBDT
In this study, we develop a feedforward deep framework to infer new LPIs. Figure 1 describes the flowchart of LPIdeepGBDT. As shown in Fig. 1, the LPIdeepGBDT framework consists of three main processes after LPI datasets are built. (1) Feature extraction. Pyfeat [38] and BioProt [39] are used to extract the original features for lncRNAs and proteins. (2) Feature selection. The lncRNA and protein features are reduced into two ddimensional vector based on dimensional reduction analysis with Principle Component Analysis (PCA). The two vectors are then concatenated to depict lncRNA–protein pairs. (3) Classification. A multiplelayer deep structure, composed of forward mapping and inverse mapping, is developed to classify lncRNA–protein pairs.
Feature extraction
Feature extraction of lncRNAs
Pyfeat [38] is widely applied to generate numerical features via sequence information. In this study, we use Pyfeat to obtain lncRNA features and represent an lncRNA as a 3, 051dimensional vector. The details are shown in Table 2.
Feature extraction of proteins
BioProt [39] utilizes various information to represent a protein. In this study, we use BioProt to obtain protein features and represent each protein as a 9,890dimensional vector. The details are shown in Table 3.
Dimension reduction
The feature dimensions of lncRNAs and protein are reduced based on PCA, respectively. Two ddimensional feature vectors are obtained and concatenated as a 2ddimensional vector \(\varvec{x}\) applied to represent an lncRNA–protein pair.
LPI prediction framework
Problem description
For a given LPI dataset \(D=(\varvec{X},\varvec{Y})\), where \((\varvec{x},\varvec{y})\) represents an lncRNA–protein pair (a training example), \(\varvec{x}\in \varvec{X}\) denotes a 2ddimensional LPI feature vector and \(\varvec{y}\in \varvec{Y}\) denotes its label, we aim to classify unknown lncRNA–protein pairs.
For a feedforward deep architecture with one original input layer, one output layer and (m1) intermediate layers, suppose that \(\varvec{o}_i\) (\(i\in \{0,1,2, \cdots , m\}\)) denotes the output in the ith layer. For an lncRNA–protein pair \((\varvec{x}, \varvec{y})\), we want to learn a mapping \({F_i}\) based on GBDT to minimize the empirical loss L between the desired output \(\varvec{y}\) and the final real output \(\varvec{o}_m\) on the training data.
Gradient boosting decision trees
GBDT can generate highly robust, interpretable and competitive classification procedures, especially for exploiting less than clean data [29, 40, 41]. For an lncRNA–protein pair \((\varvec{x},\varvec{y})\), an estimator \(f(\varvec{x})\) denotes an approximate function response to the label \(\varvec{y}\), the GBDT model iteratively builds K different individual decision tree \(\{g(\varvec{x}; \alpha _1), \dots , g(\varvec{x};\alpha _K)\}\) using the training data \(D=(\varvec{X},\varvec{Y})\). And \(f(\varvec{x})\) can be denoted as an expansion of individual decision tree \(g(\varvec{x};\alpha _k)\) by Eq. (2).
where each tree splits the input space into N disjoint regions \(\{R_{1k},\cdots , R_{jk}\}\) and calculates a constant value \(\gamma _{ik}\) for the region \(R_{jk}\) where \(I = 1\,\,if\,\,\varvec{x} \in {R_{jk}};\,I = 0,\,otherwise\). \(f_k(\varvec{x})\) denotes an addition function combined from the first decision tree to the kth decision tree. The parameters \(\alpha _{k}\) denotes the mean values of partition locations and the terminal leaf nodes for each partitioning variables in the kth decision tree. The parameters \(\beta _{k}\) denotes the weights used to determine how to effectively integrate the prediction results from individual decision trees when the leaf nodes of each collection are known. The two parameters \(\alpha _{k}\) and \(\beta _k\) can be estimated by minimizing a loss function \(L(\varvec{y},f(\varvec{x}))\) by Eq. (3).
and
To solve the model (3), Friedman [42] proposed a gradient boosting approach. First, the parameters \(\alpha _m\) can be estimated based on least square error:
where \({{{{\tilde{y}}}_{im}}}\) denotes the gradient and is defined by Eq. (6).
The parameters \(\beta _k\) can be determined by Eq. (7).
The estimator \(f_k(\varvec{x})\) for the kth regression tree can be updated by Eq. (8)
The final estimator \(f(\varvec{x})\) can be obtained by Eq. (9)
The gradient boosting approach calculates the optimal values of the parameters \(\alpha _m\) via minimizing the least square function defined by Eq. (5). The parameters \(\beta _m\) can be solved by Eqs. (5) and (7). And the GBDT algorithm is described as Algorithm 1.
The multilayered deep architecture with GBDT
We exploited a multilayered deep architecture with GBDT to classify unknown lncRNA–protein pairs. Firstly, m gradient boosting decision trees are initialized. Initial forward mapping, inverse mapping, and output are then computed. Second, pseudolabel in the mth layer is obtained based on the initialized output and real label. Third, the forward mapping for each regression tree is iteratively updated based on the computed pseudolabel at the last iteration. Fourth, the inverse mapping is iteratively learned based on the achieved forward mapping at the last iteration. Finally, the final label is output after m iterations.
Phase I: Initialize GBDT
It is very difficult to design a random tree structure based on the distribution from all potential tree configurations. Therefore, multiple Gaussian noise data are injected to the output in all intermediate layers. Given a deep structure with m layers, the initial forward mapping \(F_i^{0}\) (\(i\in \{1,2,..., m\}\)) and the inverse mapping \(G_i^{0}\) (\(i\in \{2,3, ..., m\}\)) are obtained by a few very tiny trees, where index 0 represents the tree structures achieved in the initialization procedure. In addition, the initial output \(\varvec{o}_0\) is set as X and \(\varvec{o}_i=F_i^{0}(\varvec{o}_{i1})\) (\(i\in \{1,2,..., m\}\)).
The iterations are updated based on the learned forward mappings and inverse mappings. At each iteration t, we conduct Phases IIIV.
Phase II: Compute the pseudolabel in the m th layer
The pseudolabel in the mth layer can be computed based on the final output \(\varvec{o}_m\) and the real label \(\varvec{y}\), \(\alpha\) is the learning rate by Eq. (10)
Phase III: Forward mapping
At the tth iteration, during the forward mapping, \(F_i^{t}\) for each regression tree in a GBDT is first initialized by \(F_i^{t}=F_i^{t1}\) and updated based on a pseudolabels \(\varvec{p}_{i1}^{t}\) with \(\varvec{p}_{i1}^{t}=G_i(\varvec{p}_i^{t})\). The details are described as follows.
For each regression tree in a GBDT, we define a reconstruction loss function as Eq. (11).
The pseudoresiduals for each tree can be computed by Eq. (12).
When the pseudolabel in each layer is calculated, each \(F_i^{t1}\) can implement a gradient ascent towards its pseudoresidual by Eq. (12).
Each regression tree \(g_k\) is fitted to \(\varvec{r}_k^{forw}\) based on the training set (\(\varvec{o}_{i1},\varvec{r}_k^{forw}\)) and the forward mapping \(F_i^{t}\) for each tree can be updated by Eq. (13).
Finally, we obtain the output for each layer by the forward mapping by Eq. (14).
The forward mapping procedures are described as Algorithm 2.
In this phase, we use a bottom up update technique, that is, \(F_i\) will be updated before \(F_j\) when \(i<j\). In addition, each \(F_i\) can run multiple rounds of additive boosting operations towards its current pseudolabel.
Phase IV: Inverse mapping
At the tth iteration, for each decision tree, given the forward mapping \(F_i^{t  1}\) learned from the (t2)th iteration, we intend to achieve an “pseudoinverse” mapping \(G_i^{t}\) paired with each \(F_i^{t  1}\) satisfying \(G_i^t(F_i^{t  1}({\varvec{o}_{i  1}})) \approx {\varvec{o}_{i  1}}\) based on the following expected value of the reconstructed loss function by Eq. (15):
where \(L_i^{inv}\) denotes the reconstructed loss in the ith layer.
To build a more robust and generative model, random noises \(\sigma\) are injected into the output in all intermediate layers:
For each regression tree \(g_k\) in a GBDT, the reconstructed error can be computed by Eq. (17):
Based on the noise injection, each \(G_i^{t1}\) follows a gradient ascent towards the pseudoresiduals by Eq. (18)
where \({\varvec{r}_k^{inv}}\) denotes the pseudoresiduals of the kth regression tree during the inverse mapping. For each regression tree \(g_k\) in GBDT, we fit it to \(\varvec{r}_k\) via the training set \((F_i^{t1}(\varvec{o}_{j1}^{noise}),\varvec{r}_k^{inv})\) and then update \(G_i^{t}\) by Eq. (19).
Finally, the pseudolabel in each intermediate layer can be propagated from the final layer to the first layer by Eq. (20):
For all intermediate layers and the final output layer (\(i \in \{m,m1,...,2\})\), the inverse mapping procedures are described as Algorithm 3.
We can obtain the inverse mapping \(G_i^{t}\) for the final output layer and all intermediate layers and the pseudolabels \(\varvec{p}_{i}^{t}\) for the first layer and all the intermediate layers. After finishing the tth iteration, we continue the \((t+1)\)th iteration to update \(F_i\) and \(G_i\).
During LPI prediction, a linear classifier \(\varvec{Y}=\varvec{XW}^{T}+b\) is applied to the forward mapping in the mth layer. There are two main advantages. First, the m1 layers can rerepresent the LPI features as linearly separable as possible. Second, the corresponding inverse mapping in the mth layer does not have to be computed because the pseudolabel in the (m1)th layer can be obtained based on the gradient of global loss related to the output in the (m2)th layer.
Results
The experiments is mainly explored to empirically examine if the proposed LPIdeepGBDT method can effectively predict new LPIs.
Evaluation metrics
The six measurements are utilized to evaluate the performance of LPIdeepGBDT: precision, recall, accuracy, F1score, AUC and AUPR. For the six evaluation criteria, higher values depict better performance [43]. The experiments are repeatedly implemented for 20 times. The average performance for the 20 rounds is taken as the final performance. The six measurements are defined by Eqs. (21)–(24).
where TP, TN, FP, and FN represent true positives, true negatives, false positives, and false negatives, respectively. Precision denotes the ratio of correctly predicted positive samples among all predicted positive samples. Recall represents the ratio of correctly predicted positive samples among all real positive samples. Accuracy denotes the ratio of correctly predicted positive and negative samples among all samples. F1Score is harmonic mean between precision and recall. Area Under receiver operating Characteristic Curve (AUC) is used to measure the tradeoff between TP ratio and FP ratio. Area Under PrecisionRecall curve (AUPR) is applied to evaluate the tradeoff between precision and recall.
Experimental settings
The parameters in Pyfeat are set as: kgap=5, ktuple=3, optimum=1, pseudo=1, zcurve=1, gc=1, skew=1, atgc=1, monoMono=1, monoDi=1, monoTri=1, diMono=1, diDi=1, diTri=1, triMono=1, and triDi=1. All parameters in BioProt and LPISKF are the corresponding values provided by refs. [39] and [21], respectively. The deep GBDT architecture we used is (input1616output). The parameters in the remaining methods are set the values when the corresponding methods obtain the best performance. The details are described in Table 4.
Therefore, we select two 100dimensional vectors to represent lncRNA and protein, respectively. Three 5fold Cross Validations (CVs) are carried out to evaluate the performance of LPIdeepGBDT.
5fold CV on lncRNAs (CV1): 80% of lncRNAs are extracted as train set and the remaining is test set in each round.
5fold CV on proteins (CV2): 80% of proteins are extracted as train set and the remaining is test set in each round.
5fold CV on lncRNA–protein pairs (CV3): 80% of lncRNA–protein pairs are extracted as train set and the remaining is test set in each round.
The three CVs refer to potential LPI identification for (1) a new (unknown) lncRNA without interaction information, (2) a new protein without interaction information, and (3) lncRNA–protein pairs, respectively.
Comparison with five stateoftheart LPI prediction methods
We compare the proposed LPIdeepGBDT framework with five classical LPI identification models to measure the classification performance and robustness of LPIdeepGBDT, that is, LPIBLS, LPICatBoost, PLIPCOM, LPISKF and LPIHNM. The number of negative samples is set as the same as positive samples. The best performance is illustrated in boldface in each row in Tables 5, 6 and 7.
Table 5 gives the comparative results of the five LPI identification models in terms of the six measurements under CV1. It can be observed that LPIdeepGBDT achieves better average recall, accuracy, F1score, AUC and AUPR than LPIBLS, LPICatBoost, PLIPCOM, and LPIHNM on five LPI datasets. For example, LPIdeepGBDT obtains the best average F1score value of 0.7586, 8.99%, 9.83%, 1.61%, 22.70% and 8.37% superior than LPIBLS, LPICatBoost, PLIPCOM, LPISKF, and LPIHNM, respectively. More importantly, it calculates the best AUC value of 0.8321, 1.63%, 8.32%, 2.37%, 0.02% and 6.26% better than the above five models, respectively. It also achieves the best average AUPR of 0.8095, 1.85%, 5.53%, 0.77%, 0.02% and 0.24% higher than the five methods, respectively.
LPIBLS, LPICatBoost, PLIPCOM and LPIHNM are four stateoftheart supervised learningbased LPI prediction methods and LPIdeepGBDT computes better performance than them. The results suggest the powerful classification ability of LPIdeepGBDT under CV1. More importantly, although LPIdeepGBDT computes slightly lower precision than LPISKF, other five measurements are better than LPISKF. LPISKF is one networkbased LPI inference algorithm. The type of methods have one limitation, that is, they can not be applied to predict possible interaction information for an orphan lncRNA. Therefore, LPIdeepGBDT is appropriate to prioritize underlying proteins associated with a new lncRNA.
Table 6 depicts the performance of LPIBLS, LPICatBoost, PLIPCOM, LPISKF, LPIHNM, and LPIdeepGBDT under CV2. The results show that the performance of LPIdeepGBDT is slightly lower than LPIHNM. Under CV2, 80% proteins are extracted as training set and the remaining is test set in each round. That is, there will be relatively higher proteins for which association information is masked, thereby resulting in the reduction of samples and affecting the performance of LPIdeepGBDT. Compared to other five methods, LPIHNM may be relatively robust to data abundant level when predicting possible lncRNAs for a new protein.
More importantly, LPIdeepGBDT computes the best average AUC and AUPR in comparing to LPIBLS, LPICatBoost, and PLIPCOM. For example, LPIdeepGBDT obtains the best average AUC of 0.6815, 21.97%, 9.24%, and 4.01% superior than LPIBLS, LPICatBoost, and PLIPCOM, respectively. LPIdeepGBDT achieves the best average AUPR of 0.6771, 15.74%, 10.37%, and 6.78% better than the above three methods, respectively. AUC and AUPR are two more important evaluation criteria compared to other four measurements. LPIdeepGBDT outperforms LPIBLS, LPICatBoost, and PLIPCOM in terms of AUC and AUPR. The results suggest that LPIdeepGBDT is one appropriate LPI prediction algorithm.
In particular, LPIBLS is an ensemble learningbased model. LPIdeepGBDT significantly outperforms LPIBLS based on AUC and AUPR. The results illustrate that LPIdeepGBDT may obtain better ensemble performance. In addition, LPICatBoost and PLIPCOM are two categorical boosting techniques. LPIdeepGBDT, integrating the idea of deep architecture, obtains better performance than the two methods. It shows that deep learning may more effectively learn the relevances between lncRNAs and proteins. Although LPISKF computes better AUPR than LPIdeepGBDT, LPISKF is a networkbased model. Networkbased methods can not reveal association information for an orphan protein. In summary, LPIdeepGBDT may be applied to infer possible interacting lncRNAs for a new protein.
The experimental results under CV3 are shown in Table 7. The comparative results demonstrate that LPIdeepGBDT computed the best average precision, recall, accuracy, F1score, AUC, and AUPR over all datasets. For example, LPIdeepGBDT obtains the best average F1score value of 0.8429, 14.83%, 10.77%, 3.10%, 16.73% and 18.43% superior than LPIBLS, LPICatBoost, PLIPCOM, LPISKF and LPIHNM, respectively. More importantly, it calculates the best AUC value of 0.9073, 4.93%, 11.21%, 3.32%, 0.12% and 14.49%, better than the above five models, respectively. It also achieves the best average AUPR of 0.8849, 5.82%, 8.84%, 2.59%, 2.62% and 9.13% higher than the five methods, respectively. The results characterize the superior classification performance of LPIdeepGBDT. Therefore, LPIdeepGBDT can precisely discover the potential relationships between lncRNAs and proteins based on known association information.
In addition, we investigate the performance computed by all six LPI prediction methods under the three different cross validations. The results from Tables 5, 6 and 7 show that LPIBLS, LPICatBoost, PLIPCOM, LPISKF, and LPIdeepGBDT achieve much better performance under CV3 than CV1, followed by CV2, regardless of precision, recall, accuracy, F1score, AUC or AUPR. Under CV3, cross validations are conducted on all lncRNA–protein pairs and 80% lncRNA–protein pairs are used to train the model and the remaining 20% lncRNA–protein pairs are applied to test the model. However, under CV1 or CV2, cross validations are implemented on lncRNAs or proteins, that is, 80% lncRNAs or proteins are applied to train the model and the remaining 20% lncRNAs or proteins are used to test the model. CV3 may provide more LPI information relative to CV1 and CV2. The result suggest that abundant data contribute to improve the prediction performance of LPI identification models.
Case study
In this section, we aim to mine possible association data for a new lncRNA/protein or based on known LPIs.
Identifying potential proteins for a new lncRNA
RN7SL1 is an endogenous RNA. The lncRNA is usually protected by RNAbinding protein SRP9/14. Its increase can alter the stoichiometry with SRP9/14 and thus produce unshielded RN7SL1 in stromal exosomes. After exosome transfer to breast cancer cells, unshielded RN7SL1 can activate breast cancer RIGI and promote tumor growth, metastasis, and therapy resistance [44]. Hepatocellular carcinoma patients with higher RN7SL1 concentrations also show lower survival rates. RN7SL1 may enhance hepatocellular carcinoma cell proliferation and clonogenic growth [45].
In this section, we mask all interaction information for RN7SL1 and want to infer possible proteins interacting with the lncRNA. The experiments are repeated for 10 times and the interaction probabilities between RN7SL1 and other proteins are averaged over the 10 time results. The predicted top 5 proteins interacting with RN7SL1 on human LPI datasets are described in Table 8. In Dataset 1, we can observe that RN7SL1 is predicted to interact with Q15465. Q15465 displays a cholesterol transferase and autoproteolysis activity in the reticulum endoplasmic. Its Nproduct is a morphogen required for diverse patterning events during development. It induces ventral cell fate in somites and the neural tubes. It is required for axon guidance and densely related to the anteriorposterior axis patterning in the developing limb bud [35]. In the dataset, RN7SL1 may associate with 59 proteins. In other two datasets, there does not exist any associated lncRNAs for Q15465. Although the interaction between RN7SL1 and Q15465 hasn’t been validated, among all possible associated 59 proteins, the protein is ranked as 4, 6, 8, 9, and 14 by LPICatBoost, PLIPCOM, LPISKF, LPIHNM, and LPIBLS, respectively. Therefore, the association between RN7SL1 and Q15465 need further validation.
In Dataset 2, we predict that Q13148, P07910, and Q9NZI8 may interact with RN7SL1. The interaction between Q9NZI8 and RN7SL1 is known in Dataset 3. Q13148 is a RNAbinding protein involved in various procedures in RNA biogenesis and processing. The protein controls the splicing in numerous noncoding and proteincoding RNAs, for example, proteins involved in neuronal survival and mRNAs encoding proteins related to neurodegenerative diseases. It plays important roles in maintaining mitochondrial homeostasis, mRNA stability and circadian clock periodicity, the normal skeletal muscle formation and regeneration. In Dataset 2, RN7SL1 may associate with 84 proteins. Among the 84 underlying proteins for RN7SL1, the rankings of Q13148 predicted by LPIdeepGBDT LPICatBoost, PLIPCOM, LPISKF, LPIBLS, and LPIHNM are 2, 3, 1, 3, 2, and 6, respectively. That is, all the six LPI identification models predict that there may be interaction between Q13148 and RN7SL1. Therefore, we infer that Q13148 may possibly interact with RN7SL1.
More importantly, in Dataset 2, P07910 binds to premRNA and regulates the stability and translation level of bound mRNA molecules. The protein is involved in the early procedures of spliceosome assembly and premRNA splicing. In other two human LPI datasets, there are no any known associated lncRNAs for P07910. Among 84 potential associated proteins for RN7SL1, P07910 is ranked as 3, 7, 8, 9, 11, and 9 by LPIdeepGBDT, LPIBLS, LPICatBoost, PLIPCOM, LPISKF, and LPIHNM, respectively. The ranking are relatively higher. Therefore, we predict that P07910 may associate with RN7SL1.
In Dataset 3, we observe that Q9UKV8 and Q9Y6M1 may interact with RN7SL1. The interactions between RN7SL1 and the two proteins can be retrieved in Dataset 1. That is, the predicted top 5 interaction data by LPIdeepGBDT can be validated by publications. In summary, the results from case analyses based on association prediction for a new lncRNA suggest that LPIdeepGBDT can be utilized to identify new proteins associated with a new lncRNA.
Finding potential lncRNAs interacting with a new protein
Q9UL18 is a protein required by RNAmediated gene silencing. The protein can repress the translation of mRNAs complementary to them by binding to short RNAs or short interfering RNAs. It lacks endonuclease activity and thus can cleave target mRNAs. It is still required by transcriptional gene silencing of promoter regions complementary to bound short antigene RNAs [35]. In this section, we mask the interaction information for Q9UL18 and intend to find associated lncRNAs for the protein. The predicted top 5 lncRNAs on three human LPI dataset are shown in Table 9.
In Datasets 13, Q9UL18 may interact with 935, 885, and 990 lncRNAs. It can be seen that all the predicted top 5 interactions on each dataset are validated as known LPIs. The results suggest that LPIdeepGBDT can be applied to prioritize possible lncRNAs for a new protein.
Finding new LPIs based on known LPIs
We further infer new LPIs based on LPIdeepGBDT. We rank all lncRNA–protein pairs based on the computed average interaction probabilities. Figures 2, 3, 4, 5 and 6 give the predicted 50 LPIs with the highest interaction scores. In the five figures, black dotted lines and solid lines represent unknown and known LPIs obtained from LPIdeepGBDT, respectively. Gold ovals denote proteins, deep sky blue rounded rectangles denote RNA.
There are 55,165, 74,340, 26,730, 3,815, and 71,568 known and unknown lncRNA–protein pairs on given five datasets, respectively. We observe that unknown lncRNA–protein pairs between NONHSAT023366 (RAB30AS1) and O00425, n378107 (NONHSAT007673, GAS5) and Q15717, NONHSAT143568 (LINC01572) and P35637, AthlncRNA376 (TCONS_00057930) and O22823, and ZmalncRNA530 (TCONS_00007931) and C0PLI2, which are predicted to have the highest association scores on the five datasets, are ranked as 1, 3, 1, 6, and 113, respectively.
lncRNA GAS5 has close linkages with multiple complex diseases. The lncRNA is a repressor of the glucocorticoid receptors associated with growth arrest and starvation [46]. It is downregulated in breast cancer [47]. It cam also promote microglial inflammatory response in Parkinson’s disease [48], control apoptosis in nonsmallcell lung cancer [49] and prostate cancer cell [50]. Its decreased expression indicates a poor prognosis in cervical cancer [51] and gastric cancer [52].
Q15717 increases the stability of mRNA and mediates the CDKN2A antiproliferative activity and regulates p53/TP53 expression. It increases the leptin mRNA’s stability and is involved in embryonic stem cells differentiation. In dataset 2, GAS5 have been validated to interact with P35637, and Q13148. P35637 plays an important role in diverse cellular processes including transcription regulation, DNA repair and damage response, RNA transport, and RNA splicing. It helps RNA transport, mRNA stability and synaptic homeostasis in neuronal cells. Q13148 plays a crucial role in maintaining mitochondrial homeostasis. It participates in the formation and regeneration of normal skeletal muscle, negatively regulates the expression of CDK6. The three proteins are RNAbinding proteins and have in part similar biological functions. Therefore, we infer that Q15717 may be the corresponding protein of GAS5.
Discussion and further research
lncRNAs regulate many important biological processes. They have close relationships with multiple human complex diseases. However, most of them are not annotated because of the poor evolutionary conservation. Recent researches suggest that lncRNAs implement their functions by binding to the corresponding proteins. Therefore, it is a significant work to infer potential interactions between lncRNAs and proteins. Various computational methods were designed to identify new LPIs. These models improved LPI prediction and found many potential linkages between the two entities. The predicted LPIs with higher rankings are worthy of further biomedical experimental validation.
In this manuscript, we explore an LPI identification framework (LPIdeepGBDT) based on a feedforward deep architecture with GBDTs. First, three LPI datasets and two plant datasets are retrieved. Second, the biological features of lncRNAs and proteins are selected via Pyfeat and BioProt, respectively. Third, the features are reduced based on dimensional reduction technique and concatenated to depict an lncRNA–protein pair. Finally, a multilayered deep framework is developed to find the potential relationships between the two entities. We compare LPIdeepGBDT with five classical LPI discovery methods, LPIBLS, LPICatBoost, PLIPCOM, LPISKF and LPIHNM, on the five datasets under three cross validations. The results demonstrate the superior classification ability of LPIdeepGBDT. Case studies are further implemented to conduct interaction prediction for new lncRNAs (or proteins) or based on known LPIs.
LPIdeepGBDT computes the best performance on the collected five LPI datasets. It may be in large part due to the following features. First, LPIdeepGBDT fuses multiple biological features. Second, the constructed multilayered deep framework with nondifferentiable components helps to distributedly represent the outputs in intermediate layers. Thirdly, the update procedure for each intermediate layer can reduce the global loss by updating its pseudolabel and reducing the loss in the previous layer. Finally, the random noises added in the loss function can better map the neighbor training samples to right manifold.
In the future, we will collect multiple LPI datasets from different species to better mine the relevances between lncRNAs and proteins for different species. More importantly, we will develop more effective ensemble learning model to improve the performance of LPI prediction.
Availability of data and materials
Source codes and datasets are freely available for download at https://github.com/plhhnu/LPIdeepGBDT.
Abbreviations
 LPIdeepGBDT:

Feedforward deep architecture based on gradient boosting decision trees used to discover unobserved LPIs
 LPI:

Long noncoding RNA–protein interaction
 GBDT:

Gradient boosting decision trees
 lncRNAs:

Long noncoding RNAs
 CVs:

Cross validations
References
 1.
Deng L, Wang J, Xiao Y, Wang Z, Liu H. Accurate prediction of proteinlncrna interactions by diffusion and hetesim features across heterogeneous network. BMC Bioinform. 2018;19(1):1–11.
 2.
Liu ZP. Predicting lncrnaprotein interactions by machine learning methods: a review. Curr Bioinform. 2020;15(8):831–40.
 3.
Chen X, Sun YZ, Guan NN, Qu J, Huang ZA, Zhu ZX, Li JQ. Computational models for lncrna function prediction and functional similarity calculation. Brief Funct Genom. 2019;18(1):58–82.
 4.
Chen X, Yan CC, Zhang X, You ZH. Long noncoding rnas and complex diseases: from experimental results to computational models. Brief Bioinform. 2017;18(4):558–76.
 5.
Wang, W., Dai, Q., Li, F., Xiong, Y., Wei, D.Q.: Mlcdforest: multilabel classification with deep forest in disease prediction for long noncoding rnas. Brief. Bioinform. (2020)
 6.
Zhang X, Zhou Y, Mehta KR, Danila DC, Scolavino S, Johnson SR, Klibanski A. A pituitaryderived meg3 isoform functions as a growth suppressor in tumor cells. J Clin Endocrinol Metabol. 2003;88(11):5119–26.
 7.
Pibouin L, Villaudy J, Ferbus D, Muleris M, Prospéri MT, Remvikos Y, Goubin G. Cloning of the mrna of overexpression in colon carcinoma1: a sequence overexpressed in a subset of colon carcinomas. Cancer Genet Cytogenet. 2002;133(1):55–60.
 8.
Cui, Z., Ren, S., Lu, J., Wang, F., Xu, W., Sun, Y., Wei, M., Chen, J., Gao, X., Xu, C., et al.: The prostate cancerupregulated long noncoding rna plncrna1 modulates apoptosis and proliferation through reciprocal regulation of androgen receptor. In: Urologic Oncology: Seminars and Original Investigations, vol. 31, pp. 1117–1123. Elsevier (2013)
 9.
Chen X, Yan GY. Novel human lncrnadisease association inference based on lncrna expression profiles. Bioinformatics. 2013;29(20):2617–24.
 10.
van Poppel H, Haese A, Graefen M, de la Taille A, Irani J, de Reijke T, Remzi M, Marberger M. The relationship between prostate cancer gene 3 (pca3) and prostate cancer significance. BJU Int. 2012;109(3):360–6.
 11.
Yang Z, Zhou L, Wu LM, Lai MC, Xie HY, Zhang F, Zheng SS. Overexpression of long noncoding rna hotair predicts tumor recurrence in hepatocellular carcinoma patients following liver transplantation. Ann Surg Oncol. 2011;18(5):1243–50.
 12.
Wang, W., Guan, X., Khan, M.T., Xiong, Y., Wei, D.Q.: Lmidforest: a deep forest model towards the prediction of lncrnamirna interactions. Comput. Biol. Chem. 107406 (2020)
 13.
Li Y, Sun H, Feng S, Zhang Q, Han S, Du W. Capsulelpi: a lncrnaprotein interaction predicting tool based on a capsule network. BMC Bioinform. 2021;22(1):1–19.
 14.
Li, A., Ge, M., Zhang, Y., Peng, C., Wang, M.: Predicting long noncoding rna and protein interactions using heterogeneous network model. Biomed. Res. Int. 2015 (2015)
 15.
Zhou YK, Shen ZA, Yu H, Luo T, Gao Y, Du PF. Predicting lncrnaprotein interactions with mirnas as mediators in a heterogeneous network model. Front Genet. 2020;10:1341.
 16.
Yang J, Li A, Ge M, Wang M. Relevance search for predicting lncrnaprotein interactions based on heterogeneous network. Neurocomputing. 2016;206(19):81–8.
 17.
Zhao Q, Yu H, Ming Z, Hu H, Ren G, Liu H. The bipartite network projectionrecommended algorithm for predicting long noncoding rnaprotein interactions. Mol TherapyNucleic Acids. 2018;13:464–71.
 18.
Ge M, Li A, Wang M. A bipartite networkbased method for prediction of long noncoding rnaprotein interactions. Genom Proteom Bioinform. 2016;14(1):62–71.
 19.
Xie G, Wu C, Sun Y, Fan Z, Liu J. Lpiibnra: long noncoding rnaprotein interaction prediction based on improved bipartite network recommender algorithm. Front Genet. 2019;10:343.
 20.
Zhang W, Qu Q, Zhang Y, Wang W. The linear neighborhood propagation method for predicting long noncoding rnaprotein interactions. Neurocomputing. 2018;273:526–34.
 21.
Zhou YK, Hu J, Shen ZA, Zhang WY, Du PF. Lpiskf: predicting lncrnaprotein interactions using similarity kernel fusions. Front Genet. 2020;11:1554.
 22.
Chen Y, Fu X, Li Z, Peng L, Zhuo L. Prediction of lncrnaprotein interactions via the multiple information integration. Front Bioeng Biotechnol. 2021;9:60.
 23.
Peng L, Liu F, Yang J, Liu X, Meng Y, Deng X, Peng C, Tian G, Zhou L. Probing lncrnaprotein interactions: data repositories, models, and algorithms. Front Genet. 2020;10:1346.
 24.
Liu H, Ren G, Hu H, Zhang L, Ai H, Zhang W, Zhao Q. Lpinrlmf: lncrnaprotein interaction prediction by neighborhood regularized logistic matrix factorization. Oncotarget. 2017;8(61):103975.
 25.
Zhao Q, Zhang Y, Hu H, Ren G, Zhang W, Liu H. Irwnrlpi: integrating random walk and neighborhood regularized logistic matrix factorization for lncrnaprotein interaction prediction. Front Genet. 2018;9:239.
 26.
Zhang T, Wang M, Xi J, Li A. Lpgnmf: predicting long noncoding rna and protein interaction using graph regularized nonnegative matrix factorization. IEEE/ACM Trans Comput Biol Bioinf. 2018;17(1):189–97.
 27.
Zhang W, Yue X, Tang G, Wu W, Huang F, Zhang X. Sfpellpi: sequencebased feature projection ensemble learning for predicting lncrnaprotein interactions. PLoS Comput Biol. 2018;14(12):1006616.
 28.
Fan XN, Zhang SW. Lpibls: predicting lncrnaprotein interactions with a broad learning systembased stacked ensemble classifier. Neurocomputing. 2019;370:88–93.
 29.
Deng L, Yang W, Liu H. Predprba: prediction of proteinrna binding affinity using gradient boosted regression trees. Front Genet. 2019;10:637.
 30.
Wekesa JS, Meng J, Luan Y. Multifeature fusion for deep learning to predict plant lncrnaprotein interaction. Genomics. 2020;112(5):2928–36.
 31.
Shen, Z.A., Luo, T., Zhou, Y.K., Yu, H., Du, P.F.: Npignn: predicting ncrnaprotein interactions with deep graph neural networks. Brief. Bioinform. (2021)
 32.
Feng, J., Yang, Y., Zhou, Z.H.: Multilayered gradient boosting decision trees (2018)
 33.
Xie C, Yuan J, Li H, Li M, Zhao G, Bu D, Zhu W, Wu W, Chen R, Zhao Y. Noncodev4: exploring the world of long noncoding rna genes. Nucleic Acids Res. 2014;42(D1):98–103.
 34.
Yuan J, Wu W, Xie C, Zhao G, Zhao Y, Chen R. Npinter v2. 0: an updated database of ncrna interactions. Nucleic Acids Res. 2014;42(D1):104–8.
 35.
Consortium, U.: Uniprot: a worldwide hub of protein knowledge. Nucleic Acids Res. 2019;47(D1):506–15.
 36.
Zheng X, Wang Y, Tian K, Zhou J, Guan J, Luo L, Zhou S. Fusing multiple protein–protein similarity networks to effectively predict lncrnaprotein interactions. BMC Bioinform. 2017;18(12):11–8.
 37.
Bai Y, Dai X, Ye T, Zhang P, Yan X, Gong X, Liang S, Chen M. Plncrnadb: a repository of plant lncrnas and lncrnarbp protein interactions. Curr Bioinform. 2019;14(7):621–7.
 38.
Muhammod R, Ahmed S, Md Farid D, Shatabda S, Sharma A, Dehzangi A. Pyfeat: a pythonbased effective feature generation tool for dna, rna and protein sequences. Bioinformatics. 2019;35(19):3831–3.
 39.
Márquez, B., Amaya, J.C.: Bioprot contenedor autónomo de residuos biológicos. Revista colombiana de tecnologias de avanzada 1(33) (2019)
 40.
Ding C, Wang D, Ma X, Li H. Predicting shortterm subway ridership and prioritizing its influential factors using gradient boosting decision trees. Sustainability. 2016;8(11):1100.
 41.
Shi Z, Chu Y, Zhang Y, Wang Y, Wei DQ. Prediction of blood–brain barrier permeability of compounds by fusing resampling strategies and extreme gradient boosting. IEEE Access. 2020;9:9557–66.
 42.
Friedman, J.H.: Greedy function approximation: a gradient boosting machine. Ann. Stat. 1189–1232 (2001)
 43.
Jiao Y, Du P. Performance measures in evaluating machine learning based bioinformatics predictors for classifications. Quant Biol. 2016;4(4):320–30.
 44.
Nabet BY, Qiu Y, Shabason JE, Wu TJ, Yoon T, Kim BC, Benci JL, DeMichele AM, Tchou J, Marcotrigiano J, et al. Exosome rna unshielding couples stromal activation to pattern recognition receptor signaling in cancer. Cell. 2017;170(2):352–66.
 45.
Tan C, Cao J, Chen L, Xi X, Wang S, Zhu Y, Yang L, Ma L, Wang D, Yin J, et al. Noncoding rnas serve as diagnosis and prognosis biomarkers for hepatocellular carcinoma. Clin Chem. 2019;65(7):905–15.
 46.
Kino T, Hurt DE, Ichijo T, Nader N, Chrousos GP. Noncoding rna gas5 is a growth arrestand starvationassociated repressor of the glucocorticoid receptor. Sci Signal. 2010;3(107):8–8.
 47.
MourtadaMaarabouni M, Pickard M, Hedge V, Farzaneh F, Williams G. Gas5, a nonproteincoding rna, controls apoptosis and is downregulated in breast cancer. Oncogene. 2009;28(2):195–208.
 48.
Xu W, Zhang L, Geng Y, Liu Y, Zhang N. Long noncoding rna gas5 promotes microglial inflammatory response in parkinsons disease by regulating nlrp3 pathway through sponging mir2233p. Int Immunopharmacol. 2020;85:106614.
 49.
Shi X, Sun M, Liu H, Yao Y, Kong R, Chen F, Song Y. A critical role for the long noncoding rna gas5 in proliferation and apoptosis in nonsmallcell lung cancer. Mol Carcinog. 2015;54(S1):1–12.
 50.
Pickard M, MourtadaMaarabouni M, Williams G. Long noncoding rna gas5 regulates apoptosis in prostate cancer cell lines. Biochimica et Biophysica Acta. 2013;1832(10):1613–23.
 51.
Cao S, Liu W, Li F, Zhao W, Qin C. Decreased expression of lncrna gas5 predicts a poor prognosis in cervical cancer. Int J Clin Exp Pathol. 2014;7(10):6776.
 52.
Sun M, Jin FY, Xia R, Kong R, Li JH, Xu TP, Liu YW, Zhang EB, Liu XH, De W. Decreased expression of long noncoding rna gas5 indicates a poor prognosis and promotes cell proliferation in gastric cancer. BMC Cancer. 2014;14(1):1–12.
Acknowledgements
We would like to thank all authors of the cited references.
Funding
This research was funded by the National Natural Science Foundation of China (Grant 61803151, 62072172).
Author information
Affiliations
Contributions
Conceptualization: LHP, ZW and LQZ; Funding acquisition: LHP, LQZ; Investigation: LHP and ZW; Methodology: LHP and ZW; Project administration: LHP, LQZ; Software: ZW; Validation: ZW, XFT; Writing – original draft: LHP; Writing – review and editing: LHP and ZW. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Ethics approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Zhou, L., Wang, Z., Tian, X. et al. LPIdeepGBDT: a multiplelayer deep framework based on gradient boosting decision trees for lncRNA–protein interaction identification. BMC Bioinformatics 22, 479 (2021). https://doi.org/10.1186/s12859021043998
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s12859021043998
Keywords
 lncRNA–protein interaction
 Multiplelayer deep architecture
 Gradient boosting decision tree