bioRxiv Science⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.64898/2026.10.14.759026

Integrating Multisource Target Evidence with Machine-Learning Potency Prediction for Alzheimer's Disease Drug Discovery

Pan, A.·Zhai, T.·Lee, J.·Zheng, C.·Kim, D.·Shen, L.
View original source ↗Export metadata ↓

Abstract

Alzheimer's disease (AD) target evidence is distributed across biological, chemical, structural, regulatory, and trial-registry sources, complicating reproducible target evaluation and modeling. We developed a traceable data-to-model platform that integrates AGORA-centered evidence through Target Explorer, Drug-Target Lookup, and Model Portfolio workspaces. The resource contains 967 targets, 452,249 compounds, and 1,430,260 measurements. On a curated PIK3CA dataset, target-specific fine-tuning of the Boltz-2 affinity head increased Pearson r from 0.248 to 0.560 and reduced mean absolute error from 0.900 to 0.706 on 469 held-out compounds. Separately, we screened 42 regression models over 50 random splits for ten moderate-sized target datasets; a random forest model for XDH achieved the highest mean R^2 (0.815). The platform connects curated evidence, model-ready datasets, reproducible evaluation, and public results while clearly separating contextual evidence from therapeutic claims and random-split performance from generalization to unseen chemical scaffolds.

Explore related subjects

bioinformaticsPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Pan, A., Zhai, T., Lee, J., Zheng, C., Kim, D., Shen, L.. 2026-10-20. Integrating Multisource Target Evidence with Machine-Learning Potency Prediction for Alzheimer's Disease Drug Discovery. https://doi.org/10.64898/2026.10.14.759026

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

MINT Block: multivariate integration across independent studies to identify multi-omics signatures

Integrative multi-omics analyses can identify molecular signatures spanning multiple layers of a biological system. However, signatures derived from datasets with small sample sizes may not generalise. Combining data from multiple independent studies can increase the available sample size, but may also introduce unwanted variation arising from study-specific differences. Existing approaches typically address this variation sequentially by correcting the data before downstream model fitting, which can inadvertently distort the biological signals of interest. We introduce MINT Block, a framework that integrates multi-omics data across independent studies while simultaneously accounting for study-specific unwanted variation. MINT Block supports multivariate regression and classification analyses, enabling the identification of generalisable multi-omics signatures that capture relationships between different types of molecules and predict sample phenotypes. We evaluated MINT Block using bulk and single-cell multi-omics datasets from studies of human disease and marine biology. Across these applications, we showed that MINT Block identified stable, interpretable molecular signatures and improves prediction accuracy compared with existing sequential approaches.

bioinformatics↗
bioRxiv2026

ChEBI-N - Enrichment Analysis with the ChEBI Ontology

Introduction: Biological and chemical ontologies are structured representations that can be used to map entities such as metabolites onto broader chemical and biological categories. Enrichment analysis applied to such ontologies identifies classes whose representation within a set deviates from an expected background distribution, helping to interpret complex metabolomics measurements. Objectives: This study aimed to develop a tool for ChEBI (Chemical Entities of Biological Interest) ontology-based enrichment analysis, named ChEBI-N (ChEBI eNrichment). It is a successor of the previous tool BiNChE, with new features that broaden its applicability. Methods: We reimplemented BiNChE and extended it to accept SMILES or InChI strings as well as ChEBI IDs. These broaden applicability to datasets with entities that are not yet present in ChEBI. We also constructed organism-specific background populations. Results: We demonstrate the tool on three published metabolomics datasets covering wine, cheese, and human plasma samples. ChEBI-N correctly highlighted relevant enriched classes in all three datasets, such as flavonoids in the wine dataset and fatty acid classes in the cheese dataset. Comparing organism-specific and full-ChEBI backgrounds shows that restricted backgrounds reduce the inflated significance of generic classes, although the most enriched classes remain largely consistent. Conclusion: ChEBI-N is an extended successor to BiNChE for ontology-based enrichment analysis of metabolomics data, offering a hypothesis-generating complement to existing metabolomics analysis pipelines. Availability: ChEBI-N is freely available at chebin.hastingslab.org or as a Python package (pypi.org/project/chebin), with source code at github.com/ontology-tools/chebin, archived at Zenodo (doi:10.5281/zenodo.23264219) and Software Heritage (swh:1:rev:133a41792704eb7f1db746f36a5ae97e9247b2b7).

bioinformatics↗
bioRxiv2026

geoGraph: Connectivity modelling across geographic space with hexagonal discrete global grids

1. Estimating the movement and connectivity of organisms across landscapes is central to research in ecology and evolutionary biology. Although graph-based approaches representing geographic space as weighted networks provide a flexible and computationally efficient framework for such analyses, existing tools either perform their computations only on rectangular raster grids, thus introducing systematic directional and area distortions at global scales, or rely on bespoke pipelines that limit the general applicability of recent hexagonal-grid analyses. 2. To overcome these shortcomings, here we present geoGraph, an R package that brings graph-based spatial modelling to discrete global hexagonal grids. By representing landscapes on an equal-area hexagonal tessellation of the sphere, geoGraph enables global-scale analysis without projection distortion while retaining the flexibility of the graph framework and the performance of the graph and RBGL/Boost libraries. The package integrates both raster- and vector-based GIS data into a single cost surface, supports custom grids at various resolutions, and provides a complete pipeline from raw GIS layers to graph-algorithmic outputs, together with tools for visualising and interactively editing graphs. 3. We illustrate geoGraph by revisiting previous studies of isolation by distance among human populations in the Human Genome Diversity Project panel. At a global scale, restricting movement to land raised the correlation between geographic and genetic distance from r = 0.74 (great-circle) to r = 0.84. On a finer grid within Asia, accounting for terrain ruggedness improved the fit, with the best-fitting models implying reduced movement across the Himalaya and Tibetan Plateau. These analyses demonstrate that geoGraph can model movement and connectivity appropriate to the desired spatial scale and that the level of geographic realism required is itself scale-dependent. 4. By combining discrete global grids with the power of graph algorithms, geoGraph enables seamless connectivity analysis from landscape to global scales. It is broadly applicable across landscape genetics, phylogeography, conservation planning, and studies of human dispersal, and is particularly suited to large-scale problems that rectangular-raster tools cannot address without distortion.

bioinformatics↗