Bio papers — 2026-10-12

Understanding how cultural evolution shaped human disgust avoidance is key to grasping why we are so sensitive to contaminants. Simulations of hundreds of thousands of years of evolution showed that while genetic mutations changed disgust as a trait, culture transmitted these ideas much faster than previously thought, even without direct parental modeling. This suggests that affective mechanisms like disgust sensitivity can be socially transmitted rather than purely genetic adaptations.

This idea connects to how other systems evolve; for instance, we also looked at how blood reference intervals vary globally and found no consistent geography-linked structure. This suggests that our diagnostic standards are more a product of fragmented history than universal biology, which reinforces the need for personalized frameworks rather than fixed standards.

The work on PROTACs focuses on making drug discovery better by creating smarter ways to design molecules for targeted protein degradation. Researchers developed RetroTAC, a machine learning tool that predicts how easy it will be to synthesize a PROTAC based on its molecular structure. This tool scores candidates based on synthetic depth and fragment joining, helping researchers prioritize promising molecules quickly by predicting route-derived synthesizability.

The work on understanding how species persist in ecological communities offers a general framework for assessing a species' vulnerability beyond just its sheer abundance. This framework incorporates dispersal strategies and interspecific interactions, which can be inferred from spatial abundance data so that we do not need to estimate difficult-to-measure traits like dispersal parameters to gauge how threatened a species is by being pushed out of its niche.

The results show that having more varied dispersal strategies reduces the chance of competitive exclusion for a species when its abundance is held constant. This idea connects to the work on spatiotemporal noise in communities, which suggests that disorder in space and time can stabilize diversity, providing a mechanism for how complex communities can maintain stability even with strong competitive interactions.

This concept of community structure relates to how alignment metrics fail to capture functional organization. Standard similarity measures like RSA and CKA do not distinguish between networks that share the same underlying computational circuits versus those that just produce similar outputs. This lack of distinction is a problem for mechanistic interpretability because it means we cannot tell if two neural systems are truly equivalent or just superficially similar.

Furthermore, EEG research showed that dyslexia-specific patterns in brain oscillations, particularly phase-amplitude coupling near language areas, can predict reading difficulties with high accuracy. This suggests that the way the brain couples different rhythms to speech units is a key biological mechanism underlying language processing challenges.

This finding about neural organization connects to how latent causal diffusion models attempt to map regulatory processes at a single-cell level by learning mechanistic dynamical systems of gene regulation from expression data under noise. Both lines of research are trying to move beyond simple correlation toward understanding the underlying causal or structural mechanisms that drive complex behaviors in biological and artificial systems.

The work on unifying neuronal excitability across scales is crucial because it offers a way to understand how tiny cellular mechanisms translate into the large-scale activity we observe in brains. This spatiotemporal excitatory-inhibitory model shows that temporal excitability comes from fast excitation balanced by slower inhibition, while spatial excitability emerges from short-range excitation opposed by longer-range inhibition, which recovers the lateral inhibition seen in Amari neural fields.

This framework also demonstrates that both intrinsic temporal excitability and spatial synaptic feedback are necessary for coherent mean-field events and stable spatiotemporal rhythms. Selectively tuning memductances in a specific subset of neurons can adjust the frequency of these population rhythms, showing how cellular changes influence macroscopic dynamics. This structural separation provides a physically interpretable method for analyzing neuronal excitability across different scales.

In terms of broader biological systems, the modeling of Lyme disease transmission highlights that differences in tick questing behavior and host-finding success significantly drive geographic variation in human cases between regions like the northeastern and southeastern United States. The study developed a discrete-time system to model infection dynamics, finding that while southern host communities show lower baseline infection prevalence, they can support sustained transmission with different questing behaviors.

The AI-assisted systems biology reconstruction of diabetic bone marrow dysfunction provides a foundation for understanding complex regulatory networks in this condition. Three large language models analyzed numerous publications to build a network of biological entities and interactions, which was then rigorously tested against external knowledge resources like connectomeDB2025. The analysis revealed functional modules intersecting through shared effects on endothelial function and stem cell behavior, with the endothelium acting as the principal connecting node.

Today's papers

Cultural evolution shaped human's uniquely sensitive disgust avoidance: This study suggests that cultural transmission, not just genetics, played a key role in shaping humans' heightened sensitivity to disgust. [paper]
RetroTAC: Learning Route-Derived Synthetic Accessibility for PROTACs: This paper introduces a machine learning tool called RetroTAC to predict the synthetic accessibility of PROTAC molecules based on their structure. [paper]
Similarity Analysis of Blood Count Reference Intervals Across Continents Reveals No Reproducible Population or Geography-Linked Structure and Supports Personalised Values: The analysis shows that blood reference intervals lack consistent global structure, suggesting they should be personalized rather than universal standards. [paper]
TopoLIA: Topology-Inspired Latent Interaction Attention for Drug Discovery: This paper proposes a new attention mechanism that uses topological principles to better model complex interactions in drug discovery. [paper]
Robust Machine Learning for Regulatory Sequence Modeling under Biological and Technical Distribution Shifts: This work develops a framework to test the robustness of deep learning models predicting DNA sequences against real-world biological and technical shifts. [paper]
Beyond Anxiety and Avoidance: A Dynamic Feedback-Regulation Framework for Adult Romantic Attachment: This framework proposes a control theory model to explain how momentary relational feedback regulates adult romantic attachment anxiety and avoidance. [paper]
FlyHack: Stress-Testing Mechanistic Claims in a Connectome-Constrained Visual Model: This study tests the limits of mechanistic conclusions drawn from connectome models by showing that results depend heavily on the specific model and intervention used. [paper]
Leaky protection reveals three patterns in the relative advantages of vaccination strategies across transmission intensity: This research shows how vaccine failure modes change which vaccination strategy is best depending on how intense disease transmission is. [paper]
Decoding Alignment without Encoding Alignment: A critique of similarity analysis in neuroscience: This paper argues that standard similarity metrics fail to capture the underlying functional organization of neural networks, only their aggregate outputs. [paper]
EEG Entrainment to Non-Speech Rhythmic Patterns Associated with Dimensions of Language Performance: The study finds that specific brain rhythms can predict developmental dyslexia by measuring how well brain activity entrains to speech patterns. [paper]
Dispersal diversity buffers species vulnerability to local extinction: This paper proposes a framework using competitive balance and dispersal strategies to explain how species diversity helps communities resist local extinction. [paper]
Spatiotemporal noise stabilizes unbounded diversity in strongly-competitive communities: The research demonstrates that the combination of spatial structure and environmental noise allows for stable coexistence in highly competitive ecosystems. [paper]
Oxygenation and spatial heterogeneity shape radiotherapy protocol ranking through phenotypic adaptation: This study shows that tumor response to radiation depends on how oxygen is distributed spatially, which can change the best treatment schedule. [paper]
On estimating the effective sample size of phylogenetic trees in an autocorrelated chain: This paper introduces a new method based on conditional clade distributions to accurately estimate the effective sample size for phylogenetic tree analysis. [paper]
Parameter Identifiability beyond the Fisher Information Approximation in Data-Driven Modeling: This work develops a framework to rigorously assess parameter identifiability in data-driven biological models using local derivatives. [paper]
Latent Causal Diffusions for Single-Cell Perturbation Modeling: This paper presents a generative model that uses latent causal diffusion to predict how single-cell gene expression changes when perturbed. [paper]
Excitability across scales: a unifying EI model: This model unifies cellular and population neuroscience by describing neuronal excitability through a common mixed-feedback architecture that works across different scales. [paper]
The Role of Tick Questing Behavior and Host-Finding Success in Borrelia burgdorferi Transmission in the Eastern United States: This study investigates how tick behavior and host finding success drive geographic variation in Lyme disease incidence. [paper]
Artificial Intelligence-Assisted Systems Biology Reconstruction of Diabetic Bone Marrow Dysfunction: This research uses large language models to reconstruct a regulatory network for diabetic bone marrow dysfunction, identifying key interacting biological modules. [paper]

The papers

Important terms

Cultural Evolution of Disgust: This research suggests that our sensitivity to contaminants isn't purely genetic; culture can transmit these disgust traits much faster than previously thought, even without direct parental teaching.
PROTAC Synthesis Prediction (RetroTAC): This is a machine learning tool that predicts how easy it will be to synthesize a PROTAC molecule. It scores candidates based on synthetic depth and fragment joining to help researchers prioritize promising drug designs.
Species Persistence Framework: This framework assesses species vulnerability by looking at dispersal strategies and interspecific interactions, rather than just population size, helping predict how species might be pushed out of their niche.
Functional Organization Metrics: Standard similarity measures like RSA and CKA often fail because they don't distinguish between networks that are functionally equivalent versus those that just produce similar outputs.