Computational Molecular Biology 2025, Vol.15, No.5, 245-253 http://bioscipublisher.com/index.php/cmb 24 7 make predictions on the efficiency and specificity of possible grnas (Park et al., 2015). As for off-target prediction, algorithms usually take into account situations such as mismatch, insertion and deletion one by one. Some even use machine learning or deep learning to improve judgment (Cao et al., 2025; Du et al., 2025). Many tools also come with additional features such as primer design and multi-grNA design, facilitating further verification during the experimental stage. 3.2 Comparative analysis of commonly used tools Among the numerous available tools, CHOPCHOP, CRISPOR and CRISPR-P are the most frequently used ones. Although they all do gRNA design, their respective focuses are not quite the same. CHOPCHOP is more like an integrated platform, with a friendly interface, flexible parameters and relatively intuitive output. CRISPOR is characterized by the integration of multiple scoring systems for targeting efficiency and off-target risk, and relies on extensive genomic databases for support (Manghwar et al., 2020). CRISPR-P has been specifically optimized for plant genomes, and its off-target prediction for crops such as tomatoes is more in line with reality (Naeem and Alkhnbashi, 2023). Recently, some new tools based on deep learning, such as DeepCRISPR and CCLMoff, have also begun to be used. They utilize larger datasets to improve prediction accuracy (Chuai et al., 2018). 3.3 Evaluation metrics: on-target efficiency, specificity score, and usability When evaluating whether a design algorithm is user-friendly or not, people usually first look at the targeting efficiency and specificity score. Simply put, it is about whether it can accurately cut to the target and whether it will "accidentally hurt" elsewhere (Listgarten et al., 2018). In addition to these technical indicators, researchers also attach great importance to whether the interface of the tool is intuitive, whether it runs fast, and whether it can handle large genomic data (Stemmer et al., 2015). For crop editors, these factors often directly affect the pace of experiments. Tools that can balance prediction accuracy and user experience usually help researchers design reliable grnas more smoothly, reduce unnecessary trial repetitions, and accelerate the editing process of crops such as tomatoes. 4 Criteria for Algorithm Selection in Tomato Genome Editing 4.1 Tomato genome characteristics and challenges for algorithm optimization In the genome of tomatoes, there are many repetitive sequences and complex gene families. These characteristics often make CRISPR design less troublesome because off-target problems are prone to occur. To deal with this situation, high-quality reference genomes become very important. Chromosome-level genomes of tomato genotypes such as M82 and Sweet-100 (Figure 1) (Alonge et al., 2022) provide more details for identifying similar sequences. Algorithms often need to make more detailed adjustments based on these features to avoid mistaking similar but not target sequences for targets, so as to ensure that the final editing result is accurate enough. 4.2 Integration of genome annotation data and functional gene targets When designing CRISPR targets for tomatoes, relying solely on the sequence itself is often insufficient; gene annotation information must also be taken into account. Information such as gene structure, regulatory regions, and functional domains can help determine which target sites are more "important" and more likely to affect traits such as yield and stress resistance (Chandrasekaran et al., 2021). This approach can reduce the situation where irrelevant areas are mistakenly edited and is more in line with the goals of breeding and biological research. Therefore, it is very practical when designing grnas. 4.3 Need for customization based on gene family, promoter regions, or non-coding RNAs The tomato genome contains many elements that regulate gene expression, including gene families, promoters, and various non-coding RNAs. In the face of this situation, the design of gRNA cannot be one-size-fits-all and often requires additional customization. The reason is also quite simple: these areas may be similar to each other, but their functions are different. If the design is not fine enough, it is very easy to cause off-target and even affect other gene members, thereby resulting in unwanted trait changes (Hashimoto et al., 2018). Appropriate customization can enable editors to focus more on truly relevant regulatory points and enhance the effectiveness of the entire editing project.
RkJQdWJsaXNoZXIy MjQ4ODYzNA==