Data assetization leverages the complete “Data chain” operational mechanism—from data collection, analysis, and processing to the formation of dynamic solutions—to bring about systematic changes in both internal and external environments ([33]). It provides enterprises with professional information management solutions, enhances communication efficiency, and accelerates the flow of information resources. Consequently, through the production, analysis, and use of data, it increases trading velocity in the stock market and improves investment efficiency for investors in capital markets ([31]), effectively reducing the corporate cost of equity capital.

Simultaneously, by utilizing data assets to comprehensively record production, sales, and other processes, enterprises can more accurately grasp market demand during product development and manufacturing stages, creating products that align with market trends ([35]). This provides a scientific basis for production decisions and reduces potential operational risks. Furthermore, by applying disruptive information technology innovations to thoroughly reconstruct corporate operational processes, data assetization brings new opportunities for business growth and significantly enhances performance, thereby strengthening investor confidence in the enterprise’s future development and lowering the cost of obtaining investment.

Moreover, advancing data assetization effectively prevents both underinvestment and overinvestment by enterprises, leading to increased investment returns. This effectively alleviates the risks faced by investors, thereby reducing the risk premium they demand ([36]). As a result, the corporate cost of equity capital declines accordingly. Therefore, we propose the following hypothesis:

Hypothesis 1 (H1):Dataassetizationsignificantly reduces the corporate cost of equity capital.

By transforming data into a new type of asset characterized by high transparency and high value, and endowing it with a level of information flow superiority unmatched by other assets ([13]), data assetization can significantly improve a company’s information environment. The full utilization of data enabled by data assetization promotes efficient information integration, facilitating real-time data linkage both internally and externally, as well as across internal functions. This process covers all stages from data collection and analysis to application ([16]), thereby significantly optimizing the efficiency of information flow across production, R&D, management, and innovation processes, and enhancing the transparency of various business operations. This helps investors better understand the company’s production and operational status, effectively mitigates information asymmetry between the firm and its investors, and improves the information environment. As the level of data assetization increases, management also tends to proactively convey true operational information to the capital market. This “information disclosure” improves the quality of information presented in annual reports of listed companies. Moreover, firms engaging in data assetization often attract more attention and coverage from financial media, which in turn communicates useful information about the firm’s operations to investors. Data assetization also draws the attention of professional analysts, who continuously analyze the firm’s status and improve earnings forecasts, providing specialized information to investors. Furthermore, auditors exercise greater vigilance and a higher degree of professional skepticism toward a firm’s data assets, dedicating more audit resources to identify potential material misstatement risks and issuing more prudent and rigorous audit reports, which in turn enhances investors’ ability to gain insight into the firm’s actual operational condition. In summary, data assetization can effectively improve a company’s information environment.

Based on information asymmetry theory, due to disparities in the information held by different parties, one party possesses information that the other does not have or does not fully possess, thereby creating information-advantaged and information-disadvantaged parties. In capital markets, this disparity manifests as information asymmetry both among investors and between investors and firms, which is a significant reason for the increase in corporate cost of equity capital. This is because, compared to the firm, investors do not have a full understanding of its operational status, making it difficult for them to accurately assess the value of the target investment company with insufficient information. Consequently, they face higher risks associated with holding the asset. To compensate for potential losses from such risks, investors demand a higher rate of return, which in turn increases the cost of equity capital. Additionally, stocks of firms with a poor information environment typically have higher transaction costs, contributing to a higher cost of equity capital. Conversely, if a firm has high information transparency, its cost of equity capital tends to be lower ([3]). Therefore, a poor information environment often results in uninformed investors having access to only limited information, thereby affecting expected returns and leading to an increase in the corporate cost of equity capital. Thus, improving the information environment is key to reducing the corporate cost of equity capital.

Based on this, this paper proposes hypothesis 2 as follows:

H2:Dataassetizationreduces the corporate cost of equity capital by improving the quality of the information environment.

Data assetization strongly promotes technological innovation in enterprises. On the one hand, benefiting from strong state support for the development of data technology, enterprises can actively engage in data assetization to obtain policy support and tax incentives, effectively reducing the costs associated with technological innovation and providing stable financial guarantees for their R&D investments, thereby significantly enhancing their motivation for technological innovation ([4]). On the other hand, data assetization transforms traditional innovation models. Compared with conventional innovation approaches, data helps enterprises explore technological innovation paths at lower costs and significantly improves innovation quality ([9]). In terms of innovation resources, the unique advantages formed by data assetization enable faster and more effective discovery of innovation resources, as well as a more accurate understanding of target customer information and needs, allowing for more targeted innovation activities. Furthermore, the structural characteristics of digital technologies further strengthen innovation synergies and facilitate the flow of innovation factors. This open innovation environment not only broadens the innovation perspectives of enterprises but also creates conditions for the emergence of breakthrough technological innovations. Therefore, by building an open innovation system, enterprises can continuously improve innovation performance, ultimately achieving a virtuous cycle between technological innovation and business value.

Based on dynamic capability theory, enterprises need the ability to reconfigure internal and external resources to adapt to rapidly changing market environments ([32]). Consequently, enterprises must continuously enhance their competitiveness through technological innovation to achieve sustainable survival and development. Technological innovation injects strong growth momentum into enterprises, improves future earnings expectations, not only signals to the market that the enterprise has high growth potential but also facilitates accurate assessments of the firm’s future cash flow situation, helping to ensure the stability of future earnings ([22]). From the perspective of investors, enterprises that form sustained competitive advantages through continuous technological innovation can significantly enhance market confidence in their long-term profitability. This not only makes investors more willing to invest capital in these enterprises, providing greater capital support, but also encourages the establishment of long-term, stable cooperative relationships with them. As the enterprise’s reputation and recognition in the capital market continuously improve, investors perceive such enterprises as having lower risk and higher return potential, thus being willing to provide funds at a lower cost. This means that enterprises can secure favorable terms when financing and reduce the risk premium in the financing process. Innovation leads to cost reductions, efficiency improvements, and revenue growth for enterprises, optimizing their cash flow and enhancing their intrinsic value. Based on these changes, investors reassess the equity value of the enterprise. When the equity value rises, the enterprise can raise funds on more favorable terms when issuing stocks or engaging in other forms of equity financing, thereby reducing the cost of equity capital. Based on the above analysis, we propose research hypothesis 3:

H3: Dataassetizationreduces the corporate cost of equity capital by enhancing the level of technological innovation.

Compared with traditional finance, digital finance leverages the deep integration of cutting-edge technology and financial services to expand the scope of traditional financial services, thereby accessing richer financial resources ([1]). By employing big data analytics, financial institutions can intelligently collect, accurately classify, deeply mine, and make scientific decisions based on massive amounts of data. This series of processes provides advanced technical support and diverse information supply for the development of enterprise data assetization, offering more sophisticated data collection, processing, and application capabilities. This lays a solid foundation for realizing the value of data assets ([30]), breaks down information barriers between internal and external operations, upstream and downstream supply chain participants, and banks and enterprises, and integrates digital technology with inclusive finance. By leveraging the strengths of digital technology in information collection, processing, screening, and risk identification, it enables precise matching of supply and demand among different entities ([34]). On this basis, it allows for more effective assessment of enterprise risk profiles, enhances the accuracy of investor evaluations of corporate value, improves the scientific basis and rationality of investment decisions, and reduces the required rate of return demanded by investors. Furthermore, digital finance provides high-quality technological tools for enterprise information analysis. By utilizing these advanced tools, enterprises can better formulate scientifically sound, feasible production plans and technological innovation decisions, significantly boosting investor confidence in the firm. This increased confidence is reflected in the capital market, making enterprises more attractive when raising equity capital, thereby effectively reducing the cost of equity capital and strengthening the inhibitory effect on its increase. Therefore, we propose the following:

H4:The level of digital finance development in the region where an enterprise is located positively moderates the impact of dataassetizationon the cost of equity capital.

The research model is shown in Figure 1.

Figure 1.The research model.

This paper selects A-share listed companies in China from 2011 to 2023 as the research sample. In data processing, financial listed companies, companies with *ST, ST, and PT status, and those with missing key indicators are sequentially excluded. Continuous variables are winsorized at the 1^st and 99^th percentiles, ultimately resulting in 18,793 sample observations. The financial data of listed companies are obtained from the CSMAR database, while annual report data are sourced from CNINFO. The level of digital finance development is measured using the Digital Financial Inclusion Index, with data sourced from the Peking University Internet Finance Technology Center.

3.2.1. Dependent Variable

Enterprise Equity Capital Cost(Cost)

Following the approach of Cai[8]), the equity capital cost is estimated using the MPEG model as specified in Equation (1)

Eps₁ denotes the firm’s earnings per share for the next fiscal year; Eps₂ denotes the firm’s earnings per share for the second fiscal year; p₀ represents the stock closing price at the beginning of the year; and dps₁ denotes the dividend per share for the next period, calculated as dps₁ = Eps₁*K, where K is the dividend payout ratio over the past three years.

3.2.2. Independent Variable

Enterprise DataAssetization(DA)

Following the approach of He[19]), this study measures the degree of enterprise data assetization using a text analysis method based on corporate annual reports. By reviewing policy documents such as the 14th Five-Year Plan for Digital Economy Development and the Interim Provisions on Accounting Treatment of Enterprise Data Resources, and drawing on existing literature, this study selects four words highly associated with data assets—“information,” “network,” “digital,” and “data”—as seed words. Second, corpus construction and word vector training. This study collects the full text of annual reports of all A-share listed companies from 2011 to 2023, uses Python’s jieba word segmentation library to tokenize the report texts, and removes Chinese stop words. Subsequently, the Word2Vec neural network model (Skip-gram algorithm, with window size set to 5 and vector dimension set to 200) is employed to train the corpus, obtaining distributed vector representations for each word. Third, expansion and screening of similar words. The cosine similarity between all words in the corpus and the four seed words is computed, and the top 100 words with the highest similarity are selected as candidate words. After manual review, words unrelated to the concept of data assets or those carrying negative prefixes such as “not,” “without,” or “non-” (e.g., “nonlinear,” “non-action”) are excluded, ultimately retaining 32 words closely related to data assets to form the final feature lexicon. Fourth, indicator calculation and standardization. The frequency of occurrence of the feature lexicon words in each firm’s annual report is counted (Raw_DA). To account for differences in report length, this study divides the raw frequency by the total number of pages in the annual report to obtain the average frequency per page, thereby mitigating measurement bias caused by report length. Finally, this study applies a logarithmic transformation to better capture the overall pattern of enterprise data assetization. It should be noted that the “data assetization” examined in this study, while somewhat related to the “enterprise digital transformation” discussed in existing literature, is fundamentally different in conceptual connotation. Digital transformation emphasizes the process by which enterprises utilize digital technologies (such as cloud computing, artificial intelligence, and blockchain) to reshape business processes and business models. In contrast, data assetization focuses more on the institutionalized process of recognizing data as measurable assets, enabling their participation in value creation and market transactions. In short, digital transformation highlights “using digital technologies,” whereas data assetization highlights “turning data into assets.” This distinction implies that data assetization involves not only the technological aspects of digitalization but also economic behaviors such as data rights confirmation, valuation, balance-sheet inclusion, and trading, and thus may have a more direct and far-reaching impact on firms’ financing environment.

3.2.3. Mediating Variables

1)Information Environment(FDISP)

Following the approach of Pan ([29]), this study measures the information environment using analyst forecast dispersion. The rationale is that when a firm’s public information is insufficient or its information uncertainty is high, analysts have to rely more on non-public information in their forecasts. The diversity of such information sources tends to result in greater dispersion. Therefore, a larger analyst forecast dispersion indicates that the market has less access to firm-specific information, implying a poorer information environment. Following Chu, analyst forecast dispersion is calculated as shown in Equation (2):

FDISP = SD (FEPS)/PRICE (2)

where SD (FEPS) is the standard deviation of the most recent earnings per share forecasts made by all analysts during the year, and PRICE is the firm’s stock price at the beginning of the year.

2)Technological Innovation(ITI)

This study uses R&D investment to measure the input of the firm’s technological innovation process. Technological Innovation Input = R&D Investment / Operating Revenue.

3.2.4. Moderating Variable

Digital Finance Development Level(DFI)

Following Meng ([26]), this study measures the level of digital finance development using the logarithm of the regional digital finance index compiled by the Institute of Internet Finance at Peking University.

3.2.5. Control Variables

This study selects the firm size (Size), leverage ratio (Lev), return on total assets (ROA), return on equity (ROE), operating revenue growth rate (Growth), cash flow ratio (Cashflow), CEO duality (Dual), proportion of independent directors (Indep), and shareholding ratio of the top ten shareholders (Top10) as the control variables. All variable definitions and their corresponding measurements are summarized in Table 1.

Table 1. Variable definitions and measurements.

This paper constructs a two-way fixed effects model (3) to examine the direct impact of data assetization on corporate equity capital costs.

Cost i , t represents the cost of equity capital, DA i , t denotes data assetization, the subscript i refers to the firm, the subscript t refers to the year, ∑ Industry and ∑ Year represent industry and year fixed effects, and ε i , t is the random error term.

As shown in Table 2, the mean value of enterprise data assetization (DA) is 3.3095, indicating that most enterprises have undertaken a certain degree of data assetization. The maximum value is 6.9791, while the minimum value is 0, suggesting significant variation in data assetization across firms. For enterprises with a data assetization value of 0, they have not engaged in data assetization, reflecting insufficient attention to this aspect. The mean value of the cost of equity capital (Cost) is 0.1329, with a maximum of 3.6440 and a minimum of 0.0029, indicating that the cost of equity capital varies considerably across different enterprises.

Table 2. Descriptive statistics.

The baseline regression results of the impact of data assetization on the cost of equity capital are presented in Table 3. Column (1) reports the univariate test results between enterprise data assetization and the cost of equity capital, showing that data assetization has a significant negative correlation with the cost of equity capital. Column (2) presents the impact of data assetization on the cost of equity capital after adding control variables, and the results also show a significant negative correlation. Column (3) further controls for year and industry fixed effects. The results indicate that the correlation coefficient between data assetization and the cost of equity capital is −0.0041, which is significant at the 1% level. Thus, data assetization significantly reduces the cost of equity capital.

Table 3. Baseline regression results.

Note: Values in parentheses are t-statistics; *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

4.3.1. Instrumental Variable Method

To address the issue of reverse causality, this study first employs the instrumental variable method. Since a firm’s cost of equity capital is unlikely to be affected by other firms within the same province, this study follows the approach of Niu ([27]). By using the annual provincial mean of data assetization excluding the firm itself as the instrumental variable. The results are presented in Table 4.

The Kleibergen-Paap rk LM statistic is significant, indicating that the weak instrument test is passed. The Kleibergen-Paap rk Wald F statistic is also significant, suggesting that the instrumental variable passes the under‑identification test. Column (2) of Table 3 reports the second-stage regression results of the instrumental variable method. The regression coefficient for enterprise data assetization remains significantly negative at the 1% level, confirming the robustness of the main findings. In the second stage, the coefficient for data assetization is significantly negative, further supporting Hypothesis 1.

Table 4. Endogeneity tests.

Note: Values in parentheses are t-statistics; *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively (the same as in the following tables).

4.3.2. Propensity Score Matching

This paper groups enterprises based on the mean value of data assetization: those above the mean are assigned to the treatment group, and those below the mean to the control group. First, a Logit regression is performed using control variables as covariates, followed by 1:2 nearest neighbor matching. Second, a balance test is conducted. As shown in Table 5, the standardized biases of all variables are less than 10%, and the t-values indicate no systematic differences between the treatment and control groups, confirming that the balance test is passed.

Finally, after excluding the samples that did not match, the test was re-conducted, and the results are shown in column (3) of Table 4. The regression coefficients for corporate data assets are significantly negative at the 1% level. This further indicates that corporate data assets can reduce the cost of equity capital, supporting Hypothesis H1.

Table 5. Balance test.

4.3.3. Robustness Tests

(1)High-Dimensional Fixed Effects

Considering the differences in economic development levels across regions where firms are located, this paper controls for provincial-level fixed effects based on the original model to test robustness. As shown in column (1) of Table 6, the results are significantly negative at the 1% level, supporting Hypothesis H1.

(2)Alternative Measure of the Dependent Variable

Following the approach of Chen ([10]), the OJ model (4) is used to estimate the firm’s cost of equity capital.

where A = ( g p + d p s 1 / p 0 ) / 2 ,and g 2 = ( E p s 2 − E p s 1 ) / E p s 1 ; p 0 denotes the current stock price, E p s 2 , E p s 1 represent analysts’ forecasts of earnings per share for periods t = 2 and t = 1, respectively; d p s 1 is the forecasted dividend per share for period t = 1, which is numerically equal to k ∗ E p s 1 (where k is the average dividend payout ratio over the past three years); and g p represents the constant growth rate of abnormal earnings.

The regression of data assetization on the cost of equity capital calculated using the OJ model is presented in column (3) of Table 6. The results show that data assetization is significantly negatively correlated with the cost of equity capital at the 1% level, further supporting Hypothesis H1.

Table 6. Robustness tests.

4.4.1. Mediation Mechanism

Given that the traditional stepwise mediation approach may suffer from issues such as overuse and endogeneity bias, this paper adopts the method proposed by Jiang Ting ([20]) to test the mediation mechanism. Based on the theoretical analysis above, an improved information environment can reduce the risks faced by investors, thereby lowering transaction costs and suppressing the cost of equity capital. Therefore, it is only necessary to test whether data assetization improves the information environment to verify the existence of the mediation effect. To this end, model (5) is constructed.

The specific results are shown in column (1) of Table 7, indicating that data assetization is significantly negatively correlated with analyst forecast dispersion at the 5% level. Since a smaller analyst forecast dispersion implies a better information environment for the firm, this suggests that data assetization helps improve the firm’s information environment, supporting Hypothesis H2.

Based on the theoretical analysis above, technological innovation can broaden a firm’s profit margins, improve earnings expectations, and convey positive signals that help strengthen the firm’s market position. It enhances investor confidence, facilitating access to capital support and long-term cooperation, improving the firm’s reputation in capital markets, and reducing the financing risk premium. Additionally, it can optimize cash flows, thereby affecting equity pricing and lowering the cost of equity capital. Following the approach of Jiang Ting ([20]), it is only necessary to test whether data assetization promotes technological innovation to verify the existence of the mediation effect. To this end, model (6) is constructed.

The specific results are shown in column (2) of Table 7. The regression coefficient between data assetization and R&D investment is 0.0093, which is significant at the 1% level. Since higher R&D investment indicates stronger technological innovation capability of the firm, this suggests that data assetization helps promote corporate technological innovation. Therefore, Hypothesis H3 is supported.

4.4.2. Moderating Mechanism

To examine the moderating role of the level of digital financial development, model (7) is constructed.

The specific results are shown in column (3) of Table 7. The regression coefficient of the interaction term between data assetization and the level of digital financial development on the cost of equity capital is −0.0060, which is significant at the 1% level, indicating that a higher level of digital financial development strengthens the inhibitory effect of data assetization on the cost of equity capital. Thus, Hypothesis H4 is supported.

Table 7. Mechanism testing.

To examine the heterogeneity of property rights, this paper defines a variable, nature of property rights (SOE), which takes the value of 1 for state-owned enterprises and 0 otherwise, and conducts grouped regressions. In the grouped regression presented in Table 8, the observations with SOE = 1 (5362) and those with SOE = 0 (7375) sum to 12,737 observations. There is a gap of 6056 observations compared to the full sample (18,793). This is due to: 1) following industry practice, excluding listed companies in the financial and insurance sectors (approximately 3,800 observations); 2) excluding firms whose nature of ultimate controller could not be clearly determined (approximately 800 observations); and 3) excluding approximately 1456 observations due to missing values in the dependent variable or grouping variables.

As shown in columns (1) and (2) of Table 8, data assetization exerts a significant inhibitory effect on the cost of equity capital in non-state-owned enterprises, whereas this inhibitory effect is not significant in state-owned enterprises.

Table 8. Heterogeneity analysis by nature of property rights and operating risk.

Since a firm’s operating risk is related to earnings volatility, this paper follows the approach of Li ([24]) and uses earnings volatility to measure operating risk. Specifically, the standard deviation of the rolling values of the earnings before interest, taxes, depreciation, and amortization (EBITDA) margin is used to measure the firm’s operating risk. Firms with operating risk above the median are classified as the high operating risk group, while those below the median are classified as the low operating risk group.

The specific results are shown in columns (3) and (4) of Table 8. For firms with high operating risk, the regression coefficient of data assetization on the cost of equity capital is −0.0057, which is significant at the 1% level. For firms with low operating risk, the coefficient is not significant. This may be because firms with high operating risk typically face greater future uncertainty and unstable cash flows, and often suffer from resource misallocation, leading to lower investor confidence in future operations and a higher risk premium. In this context, data assetization plays a crucial role by providing timely support. Through digital technologies, it enables accurate prediction of various risks that the firm may face, reduces operational uncertainty, helps improve resource utilization efficiency, and significantly enhances investor confidence in the firm’s future cash flows and operations. As a result, its effect on reducing the cost of equity capital is more pronounced. For firms with low operating risk, which already have stable cash flows, mature markets, sound operations, and relatively stable resource allocation, investors perceive limited risk. Although data assetization can improve operational efficiency, its impact on reducing the cost of equity capital is relatively small and unlikely to significantly affect the cost of equity capital.

In the era of the digital economy, data has become an important resource for enterprises, and data assetization can have a significant impact on capital markets. This paper takes A-share listed companies from 2011 to 2023 as the research sample and empirically examines the impact of data assetization on the cost of equity capital and its underlying mechanisms. The results show that data assetization can directly reduce the cost of equity capital. Its influence path is that data assetization suppresses the cost of equity capital by improving the information environment and enhancing technological innovation. Moreover, the level of digital financial development in the region where the firm is located can strengthen the inhibitory effect of data assetization on the cost of equity capital. Furthermore, the effect of data assetization is more pronounced in non-state-owned enterprises and firms with high operating risk.

This paper makes three contributions: It adds to micro-level data assetization research by linking it to equity cost—an overlooked angle compared to prior work on financing and growth; It broadens equity cost literature by identifying information environment and innovation as dual mechanisms, and by examining digital finance as a moderator; It unpacks heterogeneous effects by ownership and operational risk, providing actionable insights for policy and practice.

Based on the above conclusions, this paper proposes the following policy recommendations.

For enterprises, they should fully recognize the important role of data assetization in corporate development, actively explore paths to data assetization, and formulate relevant data assetization strategies by considering their own industry characteristics and business models while drawing on the experience of advanced enterprises. They should clarify the value of data assets, establish a foundational data management system, strengthen data security and compliance, and conduct regular data audits and optimization to improve data quality. Through data assetization, enterprises can build tools such as financial shared service centers and dynamic risk monitoring systems, disclose operational data in real time, enhance the quality of information disclosure, and leverage visualization technologies to make information more accessible. This enables investors to more intuitively assess profitability, thereby reducing the information risk premium. At the same time, data assetization can drive digital innovation, enabling enterprises to optimize product design through big data analytics and improve production efficiency through the Internet of Things, thereby strengthening market competitiveness and alleviating investor concerns about operational uncertainty. Additionally, enterprises should strengthen collaboration with digital financial platforms, leveraging technologies such as big data and cloud computing to enhance information transparency and bolster investor confidence, laying a solid foundation for reducing the cost of equity capital.

For the government, it should establish a special fund for data assetization and build a data asset trading platform to facilitate the circulation and sharing of data assets, thereby improving their market liquidity. At the same time, it should establish a data asset valuation and evaluation platform to provide professional references for data asset valuation, enhance market recognition of data assets, and facilitate corporate financing activities, thereby reducing the cost of equity capital. Furthermore, the government should improve policies and regulations by formulating laws and regulations related to data assets, clarifying the definition of ownership, transaction rules, and security standards. This will help regulate the market order for data assetization, provide legal guarantees for corporate data assetization, reduce risks and uncertainties in the process, stabilize investor expectations, and promote a reduction in the cost of equity capital.

Although this paper analyzes the role of data assetization in reducing the cost of equity capital, the measurement indicators for data assetization are not yet fully standardized, and differences in data management capabilities across industries and firm sizes may affect the generalizability of the findings. Future research could conduct more in-depth and detailed discussions based on industry characteristics. Additionally, while this paper considers the moderating role of digital financial development, its moderating effect may be influenced by regional financial development levels. Therefore, future research could further conduct comparative analyses across sub-regions.

Author Contributions

Conceptualization, Jianhong Tao and Xinyi Zhang; methodology, Xinyi Zhang; software, Xinyi Zhang; validation, Xinyi Zhang; formal analysis, Xinyi Zhang; investigation, Xinyi Zhang; resources, Xinyi Zhang; data curation, Xinyi Zhang; writing—original draft preparation, Xinyi Zhang; writing—review and editing, Xinyi Zhang; visualization, Xinyi Zhang; supervision, Xinyi Zhang; project administration, Xinyi Zhang; funding acquisition, Jianhong Tao. All authors have read and agreed to the published version of the manuscript.