Biology subjects

Mattsson, B.

Publications and source records attributed to Mattsson, B..

3 recordsLinked to original sources

bioRxiv2026

Identifying and Addressing Systematic Data Leakage in Protein-Ligand Affinity Benchmarks

Accurate prediction of protein-ligand binding affinity is a crucial goal in structure-based drug discovery, with the potential to significantly shorten development timelines. Recently, a new wave of machine learning models based on co-folding, such as Boltz-2 and IsoDDE, has demonstrated performance that matches or exceeds that of gold-standard physics-based methods like Free Energy Perturbation (FEP). This paper provides a critical assessment of these claims, revealing that current benchmarks are heavily influenced by data leakage, and proposes a new benchmark that explicitly controls for data leakage. We demonstrate that splitting by protein-sequence identity is inherently insufficient to prevent data leakage due to "target mirroring," in which homologous proteins with low overall sequence identity still exhibit highly correlated binding profiles. Our meta-analysis of documents in the ChEMBL 36 database identifies more than 6,000 such assay pairs and finds that leakage persists for sequence-identity thresholds as low as 0.2, well below the values commonly used in benchmarks today. Additionally, we show that a ligand-only baseline model, which lacks protein structural information, achieves surprisingly high performance on the FEP+ 4 and OpenFE benchmarks (r = 0.66 and r = 0.36, respectively). Our results indicate that current benchmarks tend to reward models for memorizing training data and exploiting localized leakage rather than truly learning biophysical principles. To address this issue, we propose the Novelty-Tiered Affinity Benchmark, in which the test data is partitioned into ligand novelty tiers. In the most challenging tier (Tanimoto similarity < 0.35), ligand-only models perform notably worse (r = 0.14), offering a clear baseline for evaluating genuine generalization. We argue that the field must move beyond sequence-based splits to ensure that AI-driven discovery translates into successful prospective laboratory research.

Molecular Biology↗

bioRxiv2020

Single cell transcriptional and functional analysis of human dopamine neurons in 3D fetal ventral midbrain organoid like cultures

Transplantation of midbrain dopamine (DA) neurons for the treatment of Parkinsons disease (PD) is a strategy that has being extensively explored and clinical trials using fetal and stem cell-derived DA neurons are ongoing. An increased understanding of the mechanisms promoting the generation of distinct subtypes of midbrain DA during normal development will be essential for guiding future efforts to precisely generate molecularly defined and subtype specific DA neurons from pluripotent stem cells. In this study, we used droplet-based scRNA-seq to transcriptionally profile a large number of fetal cells from human embryos at different stages of ventral midbrain (VM) development (6, 8, and 11 weeks post conception). This revealed that the emergence of transcriptionally distinct cellular populations was evident already at these early timepoints. To study late events of human DA differentiation and functional maturation, we established a primary fetal 3D culture system that recapitulates key molecular aspects of late human DA neurogenesis and sustains differentiation and functional maturation of DA neurons in a physiologically relevant cellular context. This approach allowed us to define the molecular identities of distinct human DA progenitors and neurons at single cell resolution and construct developmental trajectories of cell types in the developing fetal VM. Overall these findings provide a unique transcriptional profile of developing fetal VM and functionally mature human DA neurons, which can be used to quality control stem cell-derived DA neurons and guide stem cell-based therapies and disease modeling approaches in PD.

neuroscience↗

bioRxiv2020

Moose or spruce: A systems analysis model for managing conflicts between moose and forestry in Sweden

Sweden has the worlds highest density of moose (Alces alces). Moose is not only a valuable game species; it also causes forest damages and traffic accidents. To avoid moose browsing, foresters respond by planting spruce (Picea abies) to an extent that reshapes the forest landscape with impacts on both production and biodiversity. To address this problem and maintain a healthy moose population in balance with the other interests, an adaptive management based on the knowledge and experiences of local hunters and landowners is advocated. However, the different stakeholders do not agree on what is an appropriate moose population, which leads to conflicts that are hard to resolve. A key problem is that it is very difficult to encompass and foresee long-term consequences of different options for moose hunting and forest management. This makes it challenging to form coherent strategies that integrate different sectorial interests at a national level. To address this issue, we have developed a systems analysis framework for integrated modeling of the moose population, forestry, and their interactions and consequences for biodiversity. We analyze the short and long-term consequences for multiple scenarios of moose hunting and forest management. Based on the results we elucidate and quantify the trade-offs and possible synergies between moose hunting and forest production. This analysis can be used to support better informed and more constructive discussions among the stakeholders in the Swedish forest sectors, and to support policies for long term sustainable forest and moose management.

ecology↗