Figure 3: Overview of the TargetAgent pipeline. Multi-donor, multi-condition Stereo-seq sections are read at two resolutions: the cellBin view is carried through a single-cell pipeline to obtain cell types, and the bin110 view is integrated across samples for spatial domain analysis. Cross-sample integration couples spots between samples by unbalanced fused Gromov-Wasserstein optimal transport. It joins the transport couplings with a gene PPI graph to regularize the STORM irregular tensor factorization, and clusters the shared embedding into joint spatial domains with a common gene programme vocabulary. Agent panel 1 reads sequencing depth and batch confounds, cell-type composition, domain prevalence per donor, matched and unmatched transport mass, and programme factor activity to propose the disease-associated domains. Target evidence is then assembled within those domains from NEBULA differential expression, gene factor weightings with pathway annotation, and topological perturbation, via the persistent sheaf Laplacian, of one factor-charged PPI complex per disease-associated programme. Agent panel 2 reads the three axes in isolation, grounds mechanism and direction in the literature, and proposes therapeutic targets. In both panels, specialists cite only the model quantities in their own evidence packet, a critic audits every cited quantity, and a synthesis step applies a stated rule to what survives.We applied this methodology to the Stereo-seq survey of aging and Alzheimer’s disease in human prefrontal cortex \[[28](https://arxiv.org/html/#bib.bib13)\]. The study profiled post-mortem BA10 tissue from twelve male donors, six Alzheimer’s cases (ages 70–91, Braak IV–VI\[[7](https://arxiv.org/html/#bib.bib58)\], CERAD C2–3) and six age-matched controls (ages 74–95, Braak II–III, CERAD C0–1), with two cryosections per donor. The authors analyzed the array at bin110, at bin50, and at the cellBin level of individual segmented nuclei, of which 398,741 were retained after QC filtering. Cell types were called from the marker sets of Mathys et al. \[[41](https://arxiv.org/html/#bib.bib1)\] into seven classes: excitatory neurons, inhibitory neurons, astrocytes, oligodendrocytes, oligodendrocyte precursors, microglia, and endothelial cells. A spatially constrained clustering resolved seven tissue compartments: the six neocortical layers and the white matter, validated by layer markers including AQP4 in L1, HPCAL1 in L2/3, PVALB in L4, PCP4 in L5, KRT17 in L6, and MBP in white matter. The principal findings of the original study are the following. Layer architecture degrades with disease, with layers II–VI contracting and layer I expanding in severe AD alongside reactive astrocytes expressing GFAP and NEAT1. Downregulation of mRNA dominates upregulation in both the moderate and severe AD cases, and the white matter carries the largest downregulation in moderate AD, involving synaptic transcripts (SNAP25, S100B, CALM2), and myelin maintenance (MBP) and SLCO5A1. QDPR peaks in moderate AD as a compensatory response, and SPP1 and SLC38A2 are highest in moderate and lowest in severe disease. Neuroprotective transcripts that healthy aging deploys near stressed neurons, such as VGF, LMO4, and NRN1, lose that pattern in AD, and instead a heat-shock and clearance response is initiated via (HSP90AA1, HSPB1, CRYAB, SORT1). Finally, the authors nominate the transcription factor ZNF460 as a regulator of three neuronal co-expression modules that fall with disease progression. 出典: TargetAgent: Mathematical AI Co-Scientist for Drug Target Identification（arXiv）Figure 5: Target ranking and differential expression in the shortlisted domains. (a) The admitted targets from the second agentic panel, ordered by tier, with the three data axes on which a line of evidence survived the critic (dysregulation, programme membership, and topological significance) and the intervention direction fixed by the literature agent, where a grey marker denotes a target carried forward as direction-open. (b) NEBULA volcano plots for the whole-domain and cell type-stratified contrasts of AD against control in domains 3 and 5, with hits at FDR <0.05<0.05 and |log2FC|>1|\\log\_{2}\\mathrm{FC}|>1 labeled. Per-section depth enters each model as a covariate and donor section as a random intercept.Within these two domains the second panel produced a ranked target list, summarized with its supporting axes in Figure [5](https://arxiv.org/html/#S3.F5 "Figure 5 ‣ 3 Results ‣ TargetAgent: Mathematical AI Co-Scientist for Drug Target Identification")a. The candidate pool is the union over the three axes of every gene called by differential expression in a contrast with a shortlisted domain on its side of interest, every gene in the topological significance set of any programme network, and the twenty genes of largest weighting in a disease-associated factor, which gives 101 gene and domain pairs over 73 distinct genes. The panel placed 16 in a convergent tier, 13 as single-axis with a strong mechanism, 25 on hold, and 19 rejected, so that 29 targets were proposed and 54 in total were admitted for consideration; the 25 on hold were carried as direction-open because no verified literature source fixed a desired direction for treating AD, and the critic raised a disqualifying concern against 17 candidate entries. A module leads the set: NKX6-2, GPR37, UGT8, PTPRZ1, MAL, GJB1, OLIG1, ERMN, and BCAS1 are each regulators or effectors of oligodendrocyte differentiation and myelin assembly, and every one of them carries the programme and topology lines in both domains except PTPRZ1 and BCAS1, whose topology line holds in the white matter only. Within the convergent tier also sit SPP1, the one target for which the two surviving lines come from different measurement units, a donor-level differential expression call and a factor weighting, and which is assigned inhibit on the reasoning that it restrains OPC progression; the negative regulators GPR37 and PTPRZ1, assigned inhibit on the reasoning that restoring a lost process means blocking what inhibits it; and the structural myelin genes CNP, MOBP, and PLP1, which are admitted but explicitly discounted by the agents as they do not drive pathology, with MBP admitted on the programme axis alone and MAG held because its periaxonal and neurite-inhibitory roles argue opposite directions. Considering PPI topology as a statistic orthogonal to loading rankings places NKX6-2, GPR37, UGT8, MAL, and GJB1 above the structural genes. The neuronal targets are carried on single axes: VGF on its differential expression, and RGS4, CCK, and STX4 on their positions in the neuronal programme PPI network of domain 3; NRN1 is held because the valence claim behind its direction was filed without a citation. Twenty of the admitted targets were flagged by the agents as novel in this indication in the context of AD, seven among the proposed targets, NKX6-2, UGT8, GPR37, PTPRZ1, GJB1, ERMN, and BCAS1, and thirteen on hold, among them TMEM144, CLDND1, PPP1R14A, GRM3, and MID1IP1. Every literature statement behind a direction was machine-checked against the searches made during the run and against PubMed to limit LLM hallucinations \[[35](https://arxiv.org/html/#bib.bib46)\]. 出典: TargetAgent: Mathematical AI Co-Scientist for Drug Target Identification（arXiv）
