Biology subjects

Dibaeinia, P.

Publications and source records attributed to Dibaeinia, P..

5 recordsLinked to original sources

bioRxiv2026

Virtual Cells Need Context, Not Just Scale

The intersection of AI and biology has entered a phase of explosive growth, driven by the ambition to build "Virtual Cells" or computational models capable of predicting cellular responses to any perturbation. Following the success of structural biology (e.g., AlphaFold) and large language models, the field has converged on training massive, high-capacity models on large-scale single-cell data. This position paper argues that scaling model capacity is insufficient to solve the Virtual Cell problem because the primary failure mode is a lack of adequate coverage over diverse biological contexts, not insufficient model expressivity. We support this claim by reviewing recent studies showing that simple baselines perform on par with sophisticated architectures within a given biological context, and current models fail to consistently generalize across contexts. We connect this finding to the causal inference literature on transportability and contrast it with domains where scaling has succeeded. We substantiate our argument through analysis of a state-of-the-art model on a 22-million-cell immunology dataset. We conclude that the community faces a causal transport problem that cannot be solved by accumulating more data from the same distributions. Instead, we argue that contextual diversity and causal representation learning deserve increased emphasis, complementing ongoing scaling of model capacity and data volume.

genomics↗

bioRxiv2025

PRSformer: Disease Prediction from Million-Scale Individual Genotypes

Predicting disease risk from DNA presents an unprecedented emerging challenge as biobanks approach population scale sizes (N > 106 individuals) with ultra-high-dimensional features (L > 105 genotypes). Current methods, often linear and reliant on summary statistics, fail to capture complex genetic interactions and discard valuable individual-level information. We introduce PRSformer, a scalable deep learning architecture designed for end-to-end, multitask disease prediction directly from million-scale individual genotypes. PRSformer employs neighborhood attention, achieving linear O(L) complexity per layer, making Transformers tractable for genome-scale inputs. Crucially, PRSformer utilizes a stacking of these efficient attention layers, progressively increasing the effective receptive field to model local dependencies (e.g., within linkage disequilibrium blocks) before integrating information across wider genomic regions. This design, tailored for genomics, allows PRSformer to learn complex, potentially non-linear and long-range interactions directly from raw genotypes. We demonstrate PRSformers effectiveness using a unique large private cohort (N {approx} 5M) for predicting 18 autoimmune and inflammatory conditions using L {approx} 140k variants. PRSformer significantly outperforms highly optimized linear models trained on the same individual-level data and state-of-the-art summary-statistic-based methods (LDPred2) derived from the same cohort, quantifying the benefits of non-linear modeling and multitask learning at scale. Furthermore, experiments reveal that the advantage of non-linearity emerges primarily at large sample sizes (N > 1M), and that a multi-ancestry trained model improves generalization, establishing PRSformer as a new framework for deep learning in population-scale genomics.

genetics↗

bioRxiv2022

Evolution of decreased aggression in honey bees targeted brain cell gene regulatory networks

Understanding how genotypic variation results in phenotypic variation, a major challenge in biology, is especially difficult for collective behaviour because collective group phenotypes arise from complex interactions between group members1. Honeybees aggressively defend their colony from attacks with highly integrated collective behaviour in which different groups of bees play specific roles, giving rise to distinct colony-level differences in aggression. A previous genome-wide association study of a population of Africanized honeybees (Apis mellifera scutellata) from Puerto Rico that recently evolved decreased aggression identified hundreds of genes with single nucleotide polymorphisms (SNPs) that associated with colony-level variation in aggression2. Many of these SNPs also showed strong signals of selection for decreased aggression2,3, but their influence on brain function was unknown. Using brain single-cell (sc) transcriptomics and sc gene regulatory network analysis, we show here that variants of these genes give rise to genetic differences in transcription factor-target gene relationships. These differences involved the activity of several TFs, some that have been previously associated with aggression, like single stranded-binding protein c31A, and some that have been associated with tissue morphogenesis but not behaviour, like apontic. The activity of these and other TFs was located in specific brain cell populations related to olfaction and vision, the two sensory modalities that bees use in colony defence. They also implicate metabolism of serotonin, a neurochemical already known to influence honeybee aggression, but not from a genetic perspective. Surprisingly, genetic differences were more pronounced in the brains of forager bees than in similarly aged but more aggressive soldier bees, pointing to an evolutionary change in division of labour for colony defence. Our results demonstrate how group genetics can shape a collective phenotype by modulating individual brain gene regulatory network architecture.

genomics↗

bioRxiv2021

Deciphering enhancer sequence using thermodynamics-based models and convolutional neural networks

Deciphering the sequence-function relationship encoded in enhancers holds the key to interpreting non-coding variants and understanding mechanisms of transcriptomic variation. Several quantitative models exist for predicting enhancer function and underlying mechanisms; however, there has been no systematic comparison of these models characterizing their relative strengths and shortcomings. Here, we interrogated a rich data set of neuroectodermal enhancers in Drosophila, representing cis- and trans- sources of expression variation, with a suite of biophysical and machine learning models. We performed rigorous comparisons of thermodynamics-based models implementing different mechanisms of activation, repression, and cooperativity. Moreover, we developed a convolutional neural network (CNN) model, called CoNSEPT, that learns enhancer "grammar" in an unbiased manner. CoNSEPT is the first general-purpose CNN tool for predicting enhancer function in varying conditions, and we show that such complex models can suggest interpretable mechanisms. We found model-based evidence for mechanisms previously established for the studied system, including cooperative activation and short-range repression. The data also favored one hypothesized activation mechanism over another and suggested an intriguing role for a direct, distance-independent repression mechanism. Our modeling shows that while fundamentally different models can yield similar fits to data, they vary in their utility for mechanistic inference. CoNSEPT is freely available at: https://github.com/PayamDiba/CoNSEPT.

bioinformatics↗

bioRxiv2019

A single-cell expression simulator guided by gene regulatory networks

A common approach to benchmarking of single-cell transcriptomics tools is to generate synthetic data sets that resemble experimental data in their statistical properties. However, existing single-cell simulators do not incorporate known principles of transcription factor-gene regulatory interactions that underlie expression dynamics. Here we present SERGIO, a simulator of single-cell gene expression data that models the stochastic nature of transcription as well as linear and non-linear influences of multiple transcription factors on genes according to a user-provided gene regulatory network. SERGIO is capable of simulating any number of cell types in steady-state or cells differentiating to multiple fates according to a provided trajectory, reporting both unspliced and spliced transcript counts in single-cells. We show that data sets generated by SERGIO are comparable with experimental data in terms of multiple statistical measures. We also illustrate the use of SERGIO to benchmark several popular single-cell analysis tools, including GRN inference methods.

bioinformatics↗