Jeff Lotthammer

Intrinsically disordered proteins and regions (collectively IDRs) are found across all kingdoms of life and play critical roles in virtually every eukaryotic cellular process. In contrast to folded proteins, IDRs lack a stable 3D structure and are instead described in terms of a conformational ensemble, a collection of energetically accessible interconverting structures. This unique structural plasticity facilitates diverse molecular recognition and function; thus, a convenient way to view IDRs is through their ensembles. Here, we combine advances in physics-based force fields for IDPs with the power of modern multi-scale generative modeling to develop STARLING, an approach for the rapid and accurate prediction of IDR ensembles directly from sequence. STARLING enables ensembles of hundreds of conformers to be generated in seconds and works on GPUs and CPUs. This, in turn, dramatically lowers the barrier to the computational interrogation of IDR function through the lens of emergent biophysical properties in addition to traditional bioinformatic approaches. We evaluate STARLING’s accuracy against experimental data and offer a series of vignettes illustrating how STARLING can enable rapid hypothesis generation for IDR function or the interpretation of experimental data.

Protein conformational heterogeneity plays an essential role in a myriad of different biological processes. Extensive conformational heterogeneity is especially characteristic of intrinsically disordered proteins and protein regions (collectively IDRs), which lack a well-defined three-dimensional structure and instead rapidly exchange between a diverse ensemble of configurations. An emerging paradigm recognizes that the conformational biases encoded in IDR ensembles can play a central role in their biological function, necessitating understanding these sequence–ensemble relations. All-atom simulations have provided critical insight into our modern understanding of the solution behavior of IDRs. However, effectively exploring the accessible conformational space associated with large, heterogeneous ensembles is challenging. In particular, identifying poorly sampled or energetically trapped regions of disordered proteins in simulations often relies on qualitative assessment based on visual inspection of simulations and/or analysis data. These approaches, while convenient, run the risk of masking poorly sampled simulations. In this work, we present an algorithm for quantifying per-residue local conformational heterogeneity in protein simulations. Our work builds on prior work and compares the similarity between backbone dihedral angle distributions generated from molecular simulations in a limiting polymer model and across independent all-atom simulations. In this regime, the polymer model serves as a statistical reference model for extensive conformational heterogeneity in a real chain. Quantitative comparisons of probability vectors generated from these simulations reveal the extent of conformational sampling in a simulation, enabling us to distinguish between situations in which protein regions are well-sampled, poorly sampled, or folded. To demonstrate the effectiveness of this approach, we apply our algorithm to several toy, synthetic, and biological systems. Accurately assessing local conformational sampling in simulations of IDRs will help better quantify new enhanced sampling methods, ensure force field comparisons are equivalent, and provide confidence that conclusions drawn from simulations are robust.

Intrinsically disordered proteins and protein regions (collectively IDRs) are critical in numerous cellular processes. To understand how IDRs facilitate function, we need tools to accurately and rapidly identify them from sequence. While many methods for disorder prediction exist, we are currently limited by throughput and accuracy for evolutionary scale analyses. To bridge this gap, we developed metapredict V3, an updated version of our disorder predictor that enables evolutionary-scale disorder prediction. Metapredict V3 enables proteome-scale prediction with state-of-the-art accuracy in seconds and was developed with a focus on usability. It is distributed as a web server, Python software package, command-line interface, and Google Colab notebook. Here, we leverage the accuracy and throughput of metapredict V3 to predict disorder for over 20,000 proteomes to evaluate the prevalence of disorder across the kingdoms of life.

Intrinsically disordered regions (IDRs) are ubiquitous across all domains of life and play a range of functional roles. While folded domains are generally well described by a stable three-dimensional structure, IDRs exist in a collection of interconverting states known as an ensemble. This structural heterogeneity means that IDRs are largely absent from the Protein Data Bank, contributing to a lack of computational approaches to predict ensemble conformational properties from sequence. Here we combine rational sequence design, large-scale molecular simulations and deep learning to develop ALBATROSS, a deep-learning model for predicting ensemble dimensions of IDRs, including the radius of gyration, end-to-end distance, polymer-scaling exponent and ensemble asphericity, directly from sequences at a proteome-wide scale. ALBATROSS is lightweight, easy to use and accessible as both a locally installable software package and a point-and-click-style interface via Google Colab notebooks. We first demonstrate the applicability of our predictors by examining the generalizability of sequence-ensemble relationships in IDRs. Then, we leverage the high-throughput nature of ALBATROSS to characterize the sequence-specific biophysical behavior of IDRs within and between proteomes.