{"scope":{"layer":"general","tissue_scope":"unknown","species":"unknown","system":"unknown","evidence_level":"peer_reviewed","importance":"","importance_reason":""},"id":"VC0175","type":"paper","title":"Benchmarking biological knowledge-graph priors against single-cell foundation models for Perturb-seq prediction","authors_or_org":"Ni J, Zhang X, Li M, Cai Y, Xie Z, Liu Y.","date":"2026-10-09","year":"2026","venue":"Computational Biology and Chemistry","url":"https://doi.org/10.1016/j.compbiolchem.2026.109480","doi":"10.1016/j.compbiolchem.2026.109480","pmid":"42860076","brain_tumor_focus":"no","tags":["benchmark","perturb-seq","foundation-model","perturbation"],"tag_slugs":["benchmark","perturb-seq","foundation-model","perturbation"],"tag_labels":["Benchmark & validation","Perturb-seq","Foundation model","Perturbation"],"tag_labels_zh":["基准评测与验证","Perturb-seq 扰动测序","基础模型","扰动"],"unmapped_tags":[],"one_line_summary_en":"Ni et al. (Southeast University; Computational Biology and Chemistry 2026) compare STRING/Gene Ontology graph priors with dense, self-edge-only and degree-rewired supports in one fixed graph-attention backbone on four public Perturb-seq datasets (Adamson, Norman, Replogle K562 and RPE1), with GEARS, scGPT and Geneformer evaluated on the same held-out guide conditions. At 200 highly variable genes, STRING-GO improved Pearson r by +0.016 (95% interval +0.010 to +0.022) over dense support; the effect fell to +0.009 at 1000 genes. Of 36 matched contrasts with external models, 32 were within a ±0.030 envelope and four favoured the external model. The authors state the main comparison was formulated during peer-review revision and that the modest gain does not by itself establish biological relevance.","one_line_summary_zh":"Ni 等（东南大学；Computational Biology and Chemistry 2026）在同一图注意力骨架中，将 STRING/基因本体（GO）图先验与稠密、仅自环和保度重连的支撑结构对比，使用 4 个公开 Perturb-seq 数据集（Adamson、Norman、Replogle K562 与 RPE1），并在相同留出 guide 条件下评测 GEARS、scGPT 与 Geneformer。在 200 个高变基因下，STRING-GO 比稠密支撑的 Pearson r 高 +0.016（95% 区间 +0.010 至 +0.022），1000 个基因时降至 +0.009。与外部模型的 36 组匹配对比中，32 组落在 ±0.030 区间内，其余 4 组均是外部模型更优。作者说明主要对比是在审稿修改阶段提出的，且这一小幅提升本身不能说明生物学意义。","verified":"yes","verification_source":"PubMed E-utilities PMID 42860076 (title, 6 authors, Southeast University affiliations, Comput Biol Chem 126(Pt 2):109480, epub 2026-10-09, abstract) + Crossref (journal-article, issue date 2027-02) fetched 2026-10-20; full text not read","page_url":"https://gliostate.com/library/VC0175/","api_url":"https://gliostate.com/api/items/VC0175.json","sameAs":["https://doi.org/10.1016/j.compbiolchem.2026.109480","https://pubmed.ncbi.nlm.nih.gov/42860076/"],"is_preprint":false,"facets":{"type":"https://gliostate.com/library/type/paper/","year":"https://gliostate.com/library/year/2026/","tags":["https://gliostate.com/library/tag/benchmark/","https://gliostate.com/library/tag/perturb-seq/","https://gliostate.com/library/tag/foundation-model/","https://gliostate.com/library/tag/perturbation/"]},"cited_by":[],"related":[{"id":"VC0174","title":"Emerging frontiers for AI in single-cell genomics","shared_tags":["foundation-model","perturb-seq"],"page_url":"https://gliostate.com/library/VC0174/","api_url":"https://gliostate.com/api/items/VC0174.json"},{"id":"VC0143","title":"Generating high-quality Perturb-seq data for the 2026 Virtual Cell Challenge","shared_tags":["benchmark","perturb-seq"],"page_url":"https://gliostate.com/library/VC0143/","api_url":"https://gliostate.com/api/items/VC0143.json"},{"id":"VC0139","title":"Engineered regulatory states resolve context-dependent genetics in macular degeneration","shared_tags":["perturbation","perturb-seq"],"page_url":"https://gliostate.com/library/VC0139/","api_url":"https://gliostate.com/api/items/VC0139.json"},{"id":"VC0046","title":"Scalable probe-based single-cell transcriptional profiling for virtual cell perturbation mapping and synthetic biology phenotyping","shared_tags":["perturb-seq"],"page_url":"https://gliostate.com/library/VC0046/","api_url":"https://gliostate.com/api/items/VC0046.json"},{"id":"VC0045","title":"X-Atlas/Orion: Genome-wide Perturb-seq Datasets via a Scalable Fix-Cryopreserve Platform for Training Dose-Dependent Biological Foundation Models","shared_tags":["perturb-seq"],"page_url":"https://gliostate.com/library/VC0045/","api_url":"https://gliostate.com/api/items/VC0045.json"},{"id":"VC0070","title":"In vivo perturb-seq of cancer and microenvironment cells dissects oncologic drivers and radiotherapy responses in glioblastoma.","shared_tags":["perturb-seq"],"page_url":"https://gliostate.com/library/VC0070/","api_url":"https://gliostate.com/api/items/VC0070.json"}]}