HAUSP-PG: Ambiguity in Biomedical Literature

• HAUSP-PG is an ambiguous label lacking a clear definition in current biomedical and computational literature.
• Analysis indicates that HAUSP-PG is not featured as a ligand in PREP studies nor as a component in pathology-genomic graph models.
• The term likely arises from abbreviation conflation, underscoring challenges in ensuring terminological precision and accurate database curation.

HAUSP-PG does not appear as a defined ligand, computational framework, or biological entity in the cited arXiv literature. In "Decoding Binding Pathways of Ligands in Prolyl Oligopeptidase" (Walczewska-Szewc et al., 19 Mar 2025), the authors explicitly state that PG and HAUSP-PG are not discussed, while in "Pathology-genomic fusion via biologically informed cross-modality graph learning for survival analysis" (Zhang et al., 2024), the string "PG" occurs only in the context of the Pathology-Genome Heterogeneous Graph (PGHG) model. Within this evidentiary scope, HAUSP-PG is therefore best regarded as an ambiguous label rather than an established term with a stable technical meaning.

1. Terminological status in the cited literature

The strongest factual statement available is negative: the PREP ligand study identifies HUP-46, HUP-55, and KYP-2047 as analyzed ligands and then notes that neither PG nor HAUSP-PG is mentioned or analyzed (Walczewska-Szewc et al., 19 Mar 2025). The pathology-genomic study introduces PGHG as a multimodal survival-analysis framework, where "PG" abbreviates "Pathology-Genome" rather than naming a ligand or protein-associated construct (Zhang et al., 2024).

Source	Relevant term	Status
(Walczewska-Szewc et al., 19 Mar 2025)	PG; HAUSP-PG	Not mentioned or analyzed
(Zhang et al., 2024)	PG in PGHG	Abbreviation for Pathology-Genome

This usage pattern constrains interpretation. No binding site, pathway, loss function, survival endpoint, or mechanistic role can be assigned to HAUSP-PG from these sources. A plausible implication is that the label results from abbreviation conflation rather than from a formally introduced concept.

2. Relation to PREP ligand research

The PREP study is centered on prolyl oligopeptidase as a target in neurodegenerative disease and distinguishes between traditional inhibitors and HUP ligands. KYP-2047 is described as a traditional inhibitor that targets the active site between the catalytic domains of PREP and the $\beta$-propeller domain, whereas HUP ligands such as HUP-46 and HUP-55 bind alternative regions, including the hinge site, and are discussed in connection with modulation of protein-protein interactions involving pathological proteins such as $\alpha$-synuclein and Tau (Walczewska-Szewc et al., 19 Mar 2025).

The same study reports ligand-specific pathway behavior. KYP-2047 strongly prefers the central tunnel of the $\beta$-propeller domain for dissociation; HUP-46 exits through an inter-helical site; HUP-55 exhibits pathway hopping, diffusively sampling several exits before escape. Free-energy profiles are also ligand dependent: HUP-46 is associated with $\Delta F \approx -62\,\mathrm{kJ/mol}$ for the inter-helical-site pathway; HUP-55 is associated with $\Delta F \approx -33\,\mathrm{kJ/mol}$ for one inter-helical-site pathway and a low-cost hinge-to-active-site transition of $\sim -11\,\mathrm{kJ/mol}$ before an additional $\sim +30\,\mathrm{kJ/mol}$ loop-escape barrier; KYP-2047 shows a tunnel-mouth bottleneck of about $+80\,\mathrm{kJ/mol}$ with overall $\Delta F \approx -26\,\mathrm{kJ/mol}$ for tunnel exit (Walczewska-Szewc et al., 19 Mar 2025).

None of these observations can be transferred to HAUSP-PG, because the paper explicitly excludes that term. This is an important boundary condition: HAUSP-PG cannot be responsibly described as a PREP ligand, a HUP subclass member, or a molecule with hinge-site or active-site behavior on the basis of the cited PREP literature.

3. Relation to pathology-genomic graph learning

The second cited work introduces PGHG, a heterogeneous graph neural-network framework for survival analysis that integrates whole-slide images and bulk RNA-seq expression data. Its pathology subgraph uses non-overlapping WSI patches as nodes with 8-neighbor spatial adjacency, while its genomic subgraph uses statistically significant biological pathways as nodes, with edges derived from shared-gene proportions and top-10 similarity selection. Heterogeneous edges fully connect pathological nodes to pathway nodes to reflect the global character of bulk RNA-seq (Zhang et al., 2024).

PGHG further incorporates biological prior knowledge through adjacency reconstruction, gene-expression reconstruction, GSVA-score supervision for pathology feature extraction, and cross-modal embedding alignment. Node features are updated by attention-based intra- and inter-modality aggregation, global representations are obtained by attention pooling, and survival prediction is performed with gated attention fusion under a discrete-time survival objective. The model is evaluated on TCGA-LGG, FAHZU-LGG, FAHZU-GBM, and TCGA-KIRP, with C-index as the primary metric, and the paper reports that the biologically guided variant outperforms unimodal and other multimodal fusion models (Zhang et al., 2024).

Here, "PG" is semantically fixed by the paper itself: it denotes pathology-genome fusion. Accordingly, HAUSP-PG is not a variant name, module, subgraph, or interpretability component of PGHG in the cited work.

4. Sources of ambiguity and misidentification

A plausible source of confusion is the coexistence of several short labels with overlapping visual structure across unrelated domains. The PREP study uses HUP-46 and HUP-55 as names for ligands that modulate protein-protein interactions, while the survival-analysis paper uses PG as part of the acronym PGHG for pathology-genome heterogeneous graph learning. Because the PREP study also explicitly says that PG and HAUSP-PG are not discussed, any attempt to read HAUSP-PG into the HUP ligand set is unsupported (Walczewska-Szewc et al., 19 Mar 2025).

A second misconception would be to interpret the "PG" element in PGHG as a transportable suffix that can be attached to a separate biomedical term. The cited pathology-genomic paper does not license such a reading; there, PG is simply the abbreviation of "Pathology-Genome" inside a model name (Zhang et al., 2024).

A third misconception would be to assign HAUSP-PG the quantitative properties reported for analyzed objects in either paper. In the PREP context, these include pathway preference, residence behavior, and free-energy differences; in the PGHG context, they include graph-construction rules, loss terms, interpretability procedures, and C-index performance. The data block does not connect HAUSP-PG to any of them.

5. Implications for citation, indexing, and scholarly use

In a technical bibliography or knowledge graph, the safest treatment is to index the concrete entities actually defined in the cited papers: HUP-46, HUP-55, KYP-2047, PREP, and PGHG. The PREP study supplies explicit mechanistic content for the first group, including alternative-site binding, inter-helical and tunnel exits, mutational blocking effects, umbrella-sampling free-energy reconstruction, and the relation between dynamic binding and putative PPI-targeting effects (Walczewska-Szewc et al., 19 Mar 2025). The pathology-genomic study supplies explicit architectural content for PGHG, including heterogeneous graph construction, biologically guided feature learning, attention-based aggregation, and multimodal survival prediction (Zhang et al., 2024).

This suggests that cataloging HAUSP-PG as a synonym for any of those terms would be methodologically unsound. In literature review, database curation, and citation practice, undefined labels are best separated from documented nomenclature until a source explicitly establishes equivalence or introduces the term de novo.

6. Current evidentiary position

Within the cited arXiv record, HAUSP-PG has no independent definitional content. It is not a ligand in the PREP binding-pathway study, and it is not a model, module, or abbreviation in the pathology-genomic survival-analysis study. The only directly supportable conclusion is terminological: HAUSP-PG is absent as an analyzed entity, while nearby strings such as HUP-46, HUP-55, and PGHG are fully specified in their respective contexts (Walczewska-Szewc et al., 19 Mar 2025).

A plausible implication is that future clarification would require an explicit source that introduces HAUSP-PG and defines its relation, if any, to existing labels. Until such a definition is provided, the term should be treated as unresolved in this literature rather than as a recognized concept with established biochemical, computational, or clinical semantics.