readme's update

README.md

@@ -17,15 +17,13 @@ size_categories:
 
 ## Overview
 
-The Neoantigen Discovery Dataset (NDD) was developed with a clear focus on machine learning applications in neoantigen prediction. From the outset, NDD has been designed to support model development by systematically collecting a broad range of dimensions — including basic information, experimental labels, and ML-ready predictive features that can be directly used for training and validation.
-
 In neoantigen research, different strategies serve complementary roles: threshold-based approaches are effective for filtering candidates using known rules, while deep learning methods require large-scale, standardized datasets to uncover complex patterns and improve generalization.
 
 NDD addresses this need by providing a curated, standardized, and feature-rich dataset that enables the development, training, and benchmarking of new algorithms, while also facilitating integration with experimental evidence.
 
 ### Key highlights include
 
-- Integration of data from literature reports and public resources such as IEDB, TCGA, and CEDAR.  
+- Integration of data from literature reports and public resources such as IEDB, TCGA, and CEDAR. Based on the positive data, we scientifically generate corresponding negative data, which has greatly improved the accuracy of prediction.
 - Coverage of core experimental evidence alongside predicted features (binding affinity, stability, anchor mutations, driver status, etc.).  
 - A structured design that supports both peptide-level predictions and patient-level cohort analyses.  
 - Preservation of assay details to ensure reliable immunogenicity labels and reproducibility.  
@@ -38,7 +36,7 @@ NDD is not just a file collection, but a community resource for advancing model
 
 NDD entries are primarily curated from original literature reports, manually reviewed and standardized for consistency. To ensure transparency and traceability, each record includes PMID/DOI identifiers, allowing users to directly access the corresponding publications.
 
-In addition, we drew inspiration from previously published neoantigen datasets summarized by other research groups, building on their foundational efforts while extending the scope through broader standardization and feature integration.
+Meanwhile, we have also sequenced and analyzed the actual human biological data from relevant hospitals, generating valid training data for therapeutic approaches ranging from mutation to immunogenicity and clinical results.
 
 The curation process draws on the frameworks established by public databases such as IEDB and CEDAR, while also integrating cross-references to resources like TCGA. This approach ensures that NDD combines the breadth of literature-derived evidence with the rigor of structured annotation.