-
Characterizing control between interacting subsystems with deep Jacobian estimation
Authors:
Adam J. Eisen,
Mitchell Ostrow,
Sarthak Chandra,
Leo Kozachkov,
Earl K. Miller,
Ila R. Fiete
Abstract:
Biological function arises through the dynamical interactions of multiple subsystems, including those between brain areas, within gene regulatory networks, and more. A common approach to understanding these systems is to model the dynamics of each subsystem and characterize communication between them. An alternative approach is through the lens of control theory: how the subsystems control one ano…
▽ More
Biological function arises through the dynamical interactions of multiple subsystems, including those between brain areas, within gene regulatory networks, and more. A common approach to understanding these systems is to model the dynamics of each subsystem and characterize communication between them. An alternative approach is through the lens of control theory: how the subsystems control one another. This approach involves inferring the directionality, strength, and contextual modulation of control between subsystems. However, methods for understanding subsystem control are typically linear and cannot adequately describe the rich contextual effects enabled by nonlinear complex systems. To bridge this gap, we devise a data-driven nonlinear control-theoretic framework to characterize subsystem interactions via the Jacobian of the dynamics. We address the challenge of learning Jacobians from time-series data by proposing the JacobianODE, a deep learning method that leverages properties of the Jacobian to directly estimate it for arbitrary dynamical systems from data alone. We show that JacobianODEs outperform existing Jacobian estimation methods on challenging systems, including high-dimensional chaos. Applying our approach to a multi-area recurrent neural network (RNN) trained on a working memory selection task, we show that the "sensory" area gains greater control over the "cognitive" area over learning. Furthermore, we leverage the JacobianODE to directly control the trained RNN, enabling precise manipulation of its behavior. Our work lays the foundation for a theoretically grounded and data-driven understanding of interactions among biological subsystems.
△ Less
Submitted 2 July, 2025;
originally announced July 2025.
-
Beyond Geometry: Comparing the Temporal Structure of Computation in Neural Circuits with Dynamical Similarity Analysis
Authors:
Mitchell Ostrow,
Adam Eisen,
Leo Kozachkov,
Ila Fiete
Abstract:
How can we tell whether two neural networks utilize the same internal processes for a particular computation? This question is pertinent for multiple subfields of neuroscience and machine learning, including neuroAI, mechanistic interpretability, and brain-machine interfaces. Standard approaches for comparing neural networks focus on the spatial geometry of latent states. Yet in recurrent networks…
▽ More
How can we tell whether two neural networks utilize the same internal processes for a particular computation? This question is pertinent for multiple subfields of neuroscience and machine learning, including neuroAI, mechanistic interpretability, and brain-machine interfaces. Standard approaches for comparing neural networks focus on the spatial geometry of latent states. Yet in recurrent networks, computations are implemented at the level of dynamics, and two networks performing the same computation with equivalent dynamics need not exhibit the same geometry. To bridge this gap, we introduce a novel similarity metric that compares two systems at the level of their dynamics, called Dynamical Similarity Analysis (DSA). Our method incorporates two components: Using recent advances in data-driven dynamical systems theory, we learn a high-dimensional linear system that accurately captures core features of the original nonlinear dynamics. Next, we compare different systems passed through this embedding using a novel extension of Procrustes Analysis that accounts for how vector fields change under orthogonal transformation. In four case studies, we demonstrate that our method disentangles conjugate and non-conjugate recurrent neural networks (RNNs), while geometric methods fall short. We additionally show that our method can distinguish learning rules in an unsupervised manner. Our method opens the door to comparative analyses of the essential temporal structure of computation in neural circuits.
△ Less
Submitted 29 October, 2023; v1 submitted 16 June, 2023;
originally announced June 2023.
-
A Lattice Model of Charge-Pattern-Dependent Polyampholyte Phase Separation
Authors:
Suman Das,
Adam Eisen,
Yi-Hsuan Lin,
Hue Sun Chan
Abstract:
In view of recent intense experimental and theoretical interests in the biophysics of liquid-liquid phase separation (LLPS) of intrinsically disordered proteins (IDPs), heteropolymer models with chain molecules configured as self-avoiding walks on the simple cubic lattice are constructed to study how phase behaviors depend on the sequence of monomers along the chains. To address pertinent general…
▽ More
In view of recent intense experimental and theoretical interests in the biophysics of liquid-liquid phase separation (LLPS) of intrinsically disordered proteins (IDPs), heteropolymer models with chain molecules configured as self-avoiding walks on the simple cubic lattice are constructed to study how phase behaviors depend on the sequence of monomers along the chains. To address pertinent general principles, we focus primarily on two fully charged 50-monomer sequences with significantly different charge patterns. Each monomer in our models occupies a single lattice site and all monomers interact via a screened pairwise Coulomb potential. Phase diagrams are obtained by extensive Monte Carlo sampling performed at multiple temperatures on ensembles of 300 chains in boxes of sizes ranging from $52\times 52\times 52$ to $246\times 246\times 246$ to simulate a large number of different systems with the overall polymer volume fraction $φ$ in each system varying from $0.001$ to $0.1$. Phase separation in the model systems is characterized by the emergence of a large cluster connected by inter-monomer nearest-neighbor lattice contacts and by large fluctuations in local polymer density. The simulated critical temperatures, $T_{\rm cr}$, of phase separation for the two sequences differ significantly, whereby the sequence with a more "blocky" charge pattern exhibits a substantially higher propensity to phase separate. The trend is consistent with our sequence-specific random-phase-approximation (RPA) polymer theory, but the variation of the simulated $T_{\rm cr}$ with a previously proposed "sequence charge decoration" pattern parameter is milder than that predicted by RPA. Ramifications of our findings for the development of analytical theory and simulation protocols of IDP LLPS are discussed.
△ Less
Submitted 2 March, 2018;
originally announced March 2018.
-
TreeOTU: Operational Taxonomic Unit Classification Based on Phylogenetic Trees
Authors:
Dongying Wu,
Ladan Doroud,
Jonathan A. Eisen
Abstract:
Our current understanding of the taxonomic and phylogenetic diversity of cellular organisms, especially the bacteria and archaea, is mostly based upon studies of sequences of the small- subunit rRNAs (ssu-rRNAs). To address the limitation of ssu-rRNA as a phylogenetic marker, such as copy number variation among organisms and complications introduced by horizontal gene transfer, convergent evolutio…
▽ More
Our current understanding of the taxonomic and phylogenetic diversity of cellular organisms, especially the bacteria and archaea, is mostly based upon studies of sequences of the small- subunit rRNAs (ssu-rRNAs). To address the limitation of ssu-rRNA as a phylogenetic marker, such as copy number variation among organisms and complications introduced by horizontal gene transfer, convergent evolution, or evolution rate variations, we have identified protein- coding gene families as alternative Phylogenetic and Phylogenetic Ecology markers (PhyEco). Current nucleotide sequence similarity based Operational Taxonomic Unit (OTU) classification methods are not readily applicable to amino acid sequences of PhyEco markers. We report here the development of TreeOTU, a phylogenetic tree structure based OTU classification method that takes into account of differences in rates of evolution between taxa and between genes. OTU sets built by TreeOTU are more faithful to phylogenetic tree structures than sequence clustering (non phylogenetic) methods for ssu-rRNAs. OTUs built from phylogenetic trees of protein coding PhyEco markers are comparable to our current taxonomic classification at different levels. With the included OTU comparing tools, the TreeOTU is robust in phylogenetic referencing with different phylogenetic markers and trees.
△ Less
Submitted 28 August, 2013;
originally announced August 2013.
-
Systematic identification of gene families for use as markers for phylogenetic and phylogeny- driven ecological studies of bacteria and archaea and their major subgroups
Authors:
Dongying Wu,
Guillaume Jospin,
Jonathan A. Eisen
Abstract:
With the astonishing rate that the genomic and metagenomic sequence data sets are accumulating, there are many reasons to constrain the data analyses. One approach to such constrained analyses is to focus on select subsets of gene families that are particularly well suited for the tasks at hand. Such gene families have generally been referred to as marker genes. We are particularly interested in i…
▽ More
With the astonishing rate that the genomic and metagenomic sequence data sets are accumulating, there are many reasons to constrain the data analyses. One approach to such constrained analyses is to focus on select subsets of gene families that are particularly well suited for the tasks at hand. Such gene families have generally been referred to as marker genes. We are particularly interested in identifying and using such marker genes for phylogenetic and phylogeny-driven ecological studies of microbes and their communities. We therefore refer to these as PhyEco (for phylogenetic and phylogenetic ecology) markers. The dual use of these PhyEco markers means that we needed to develop and apply a set of somewhat novel criteria for identification of the best candidates for such markers. The criteria we focused on included universality across the taxa of interest, ability to be used to produce robust phylogenetic trees that reflect as much as possible the evolution of the species from which the genes come, and low variation in copy number across taxa. We describe here an automated protocol for identifying potential PhyEco markers from a set of complete genome sequences. The protocol combines rapid searching, clustering and phylogenetic tree building algorithms to generate protein families that meet the criteria listed above. We report here the identification of PhyEco markers for different taxonomic levels including 40 for all bacteria and archaea, 114 for all bacteria, and much more for some of the individual phyla of bacteria. This new list of PhyEco markers should allow much more detailed automated phylogenetic and phylogenetic ecology analyses of these groups than possible previously.
△ Less
Submitted 2 July, 2013;
originally announced July 2013.
