Pith. sign in

REVIEW 2 major objections 4 minor 42 references

Hierarchical cell identities emerge from animal gene regulatory mechanisms

T0 review · 2 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read The paper claims cell-identity hierarchies emerge from enhancer competition for shared epigenetic readers, proved via a coarse-graining theorem making weighted averages of terminal expression programs stable progenitor states at…

desk verdict The central symmetry reduction is unproven, so the persistence-length predictions rest on an unstated parameter assumption; but the coarse-graining idea deserves a referee. read the letter →

arxiv 2412.11336 v3 pith:M7P56L4G submitted 2024-12-15 q-bio.CB q-bio.MN

classification q-bio.CBq-bio.MN MSC 92B0592C3737N25
keywords hierarchicalcellidentityenhancercompetitionepigeneticreadersprogenitorstatesmultilineageprimingcoarse-grainingtransformationhematopoiesisdifferentiationtherapy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Hierarchical cell identity — stem, progenitor, terminal — is a signature of animal development, but this paper argues the hierarchy is not wired into the gene-regulatory network; it emerges from how animal enhancers work. The mechanism is enhancer competition: enhancers recruit epigenetic writers that acetylate their histones, and acetylated enhancers then compete for limiting epigenetic reader molecules (such as Brd4) that drive transcription. Because cell-identity enhancers are co-bound by the very genes they activate, the network dynamics reduce to a symmetric 'gradient' form, and the paper proves a coarse-graining theorem: at intermediate competition strength (a parameter $\beta$ set by the writer/eraser balance), softmax-weighted averages of terminal expression programs, $\Xi^*_L = \sum_{i \in L} \Xi_i e^{w_i}/\sum_{i \in L} e^{w_i}$, become stable attractors — progenitor states — with persistence length of order $\mu_L^{-2}$, where $\mu_L$ is the widest angle among the averaged programs. If correct, this one principle reconstructs the observed blood hierarchy: ranking all subsets of the eleven mouse blood lineage programs by persistence length puts the known progenitors on top, explains multilineage priming and signaling biases such as erythropoietin's, and predicts why HDAC inhibitors (which raise $\beta$) drive cancer cells to differentiate.

What carries the argument

The load-bearing object is the coarse-graining transformation of Eq. 7: for a subset $L$ of enhancer types with weights $w_i$ and patterns $\Xi_i$, the transformed pattern $\Xi^*_L = \sum_{i\in L} \Xi_i e^{w_i} / \sum_{i\in L} e^{w_i}$ (a softmax-weighted average) together with $w^*_L = \log \sum_{k \in L} e^{w_k}$ rewrites the network so that its dynamics match the original fine-grained dynamics to $O(\beta \mu_L^2)$. This transformation is exact when the averaged patterns are identical — the autoregulatory-motif case that justifies the symmetric gradient form $\dot{x} = \Xi^T \mathrm{softmax}(\beta \Xi x + w) - x = -\nabla V$ with $V = -\beta^{-1}\log Z + \tfrac{1}{2}\|x\|^2$ — and it is commutative and self-similar under repeated application. It converts the biology of enhancer competition and cooperation into a quantitative statement: a subset of expression programs becomes a stable progenitor exactly when $\beta$ drops below a threshold of order $\mu_L^{-2}$, which is the 'persistence length' used to predict which progenitors exist.

What would settle it

Compute the persistence length $1/\mu_L^2$ for every subset of the eleven blood lineage programs from an independent, high-resolution expression dataset, then check the ranking claim: every documented progenitor should sit in the top tier of its size class, and every top-ranked subset should be findable by lineage tracing; one observed progenitor with a short persistence length, or one high-ranked subset that exhaustive tracing fails to find, would break the central prediction. A second, independent check targets the $\beta$ axis: the model predicts that graded HDAC inhibition (raising $\beta$) should progressively destabilize progenitors and push the population toward terminal states, so a titration experiment in a tissue with known terminal programs that instead freezes or stabilizes an intermediate state would contradict the writer/eraser control claim.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that the enhancer-regulated dynamics $\dot{x} = Q^T \mathrm{softmax}(\beta \Xi x + w) - x$ are self-similar under a specific coarse-graining. For any subset $L$ of enhancer types, replacing the patterns $\Xi_i$, $i \in L$, by the weighted average $\Xi^*_L = \sum_{i\in L} \Xi_i e^{w_i} / \sum_{i\in L} e^{w_i}$ and the weights by $w^*_L = \log \sum_{i \in L} e^{w_i}$ reproduces the original dynamics up to a correction of order $\beta \mu_L^2$, where $\mu_L$ is the maximal angle between patterns in $L$; the transformed potential satisfies $V_L \approx V + O(\beta \mu_L^2)$. Hence $\Xi^*_L$, the 'progenitor pattern', is a stable attractor of the fine-grained network at intermediate $\beta$, with persistence length $\sim \mu_L^{-2}$; for $|L|$ isolated equal-angle patterns the destabilization threshold is $\beta_{\mathrm{crit}} \approx 2|L| \mu^{-2}$. The authors derive this for the symmetric (gradient) form obtained when autoregulatory enhancer motifs make $Q \approx \Xi$, and they show the resulting hierarchy is jointly controlled by $\beta$ and by the activation vector $w$: raising $w_i$ enlarges pattern $i$'s basin and pulls all its associated progenitors toward it, modularly. Applied to hematopoiesis, persistence lengths computed from the eleven mouse blood lineage programs rank the observed progenitor states above competing subsets, including recently identified progenitors, and annealing simulations with noise and production feedback reproduce balanced blood output that can be biased, as erythropoietin biases toward erythrocytes.

Load-bearing premise

The assumption the whole argument rests on is that, once the enhancers that are always activated together are merged, the network is effectively symmetric — that is, the binding matrix that says which transcription factors switch on an enhancer is the same as the matrix that says how strongly that enhancer drives each factor's expression, so the dynamics can be written as downhill motion on a single energy landscape.

Editorial extensions

If this is right

  • In blood formation only the subsets of the eleven lineage programs with the longest persistence lengths should appear as stable progenitors; the paper verifies this ranking against the known catalog of mouse blood progenitors, including the recently identified basophil/megakaryocyte/erythrocyte and eosinophil/neutrophil states.
  • Enhancer competition becomes a doseable control axis: lowering $\beta$ stabilizes progenitor states while raising $\beta$ drives differentiation, which is the paper's mechanism for why HDAC inhibitors act as differentiation therapy and why mutations that blunt $\beta$ or widen program overlap cause blocked maturation.
  • Transcription-factor mutations that make two lineage programs more similar shrink $\mu_L$ and lengthen the persistence of the shared progenitor, explaining how factors such as Lmo2, Tal1, and Erg stabilize mixed stem-differentiated leukemia states.
  • Signaling inputs $w_i$ bias differentiation modularly: raising one terminal program's weight enlarges its basin and shifts every progenitor containing it toward that fate, without touching unrelated progenitors — the quantitative form of erythropoietin's effect on erythroid output.
  • Annealing protocols (lower then slowly raise $\beta$), combined with noise and feedback on $w$, yield controlled transitions between identity states and balanced production across fates, giving a general recipe for steering differentiation in any tissue with known terminal programs.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The persistence-length rule is derived for the dynamics in general, not for blood specifically, so the same ranking calculation should transfer to other well-mapped hierarchies such as neural crest, thymic epithelium, or intestinal lineages; testing it there is a direct way to probe the claim's generality beyond hematopoiesis.
  • Because the progenitor pattern is a softmax-weighted average with weights $e^{w_i}$, progenitor identity is graded in the signaling level $w_i$ rather than discrete; a 'bipotential' state is a continuum of weighted mixtures, which may explain why lineage-tracing studies disagree about which progenitors genuinely exist — a reading the paper gestures at but does not develop.
  • The same machinery could be inverted as a discovery tool: terminal expression programs could be treated as stored patterns and the coarse-graining transformation used to generate candidate progenitor profiles for single-cell validation in tissues where progenitors are not yet catalogued.
  • The symmetry assumption $Q \approx \Xi$ is presented as an approximation; if it fails in real networks, the persistence-length ordering might persist with rescaled effective $\beta$, but the paper does not test this robustness, so whether the hierarchy prediction survives asymmetric enhancer coupling remains open.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper proposes that hierarchical cell identity arises as an emergent property of enhancer-based gene regulation in animals. Starting from a biophysical model of enhancer acetylation and competition for shared epigenetic readers, the authors write the general dynamics in Eq. (4) and then approximate them by the symmetric gradient system Eq. (5). They define a coarse-graining transformation Eq. (7) that maps sets of terminal expression patterns to weighted-average 'progenitor patterns,' and they argue that these progenitor patterns are stable attractors over a persistence length scale set by the maximal angle between patterns, β_L ≈ μ_L^{-2}. The framework is then applied to hematopoiesis, where observed progenitor states are claimed to be those with longest persistence lengths, to erythropoietin-mediated fate bias, to cancer-related transcription factor dysregulation, and to differentiation therapy with HDAC inhibitors. The paper also draws a formal analogy between the symmetric dynamics and dense associative memory networks.

Significance. If the central reduction from Eq. (4) to Eq. (5) can be justified, the paper offers a mechanistically grounded and unusually predictive account of multilineage priming: it yields a quantitative ranking of progenitor states from terminal expression patterns, recovers known blood progenitors including recently identified ones, and makes falsifiable predictions about β modulation by HDAC/HAT perturbations. The mathematical derivations in Supplement F (equal-angle β_crit) and Supplement B (coarse-graining transformation) are self-contained and clean, and the paper ships simulation code and uses external Haemopedia expression data for the blood progenitor ranking. The connection to dense associative memories is original and potentially generative. The main risk is that the entire attractor/persistence-length machinery is derived for Eq. (5), so the validity of the Q ≈ Ξ reduction determines whether the biological conclusions follow.

major comments (2)
  1. [Model for enhancer-regulated gene expression, Eq. (4)–(5); Supplement B.3] The reduction from the general dynamics Eq. (4) to the symmetric gradient dynamics Eq. (5) is not actually derived. In the autoregulatory motif specified in Supplement B.3, each enhancer drives a distinct TF, Q_i = e_i, while the binding rows are identical, Ξ_i = z. Applying the coarse-graining transformation Eq. (7) gives Q*_L = Σ_{i∈L} e_i e^{w_i}/Σ_{i∈L} e^{w_i} = softmax(w_i), whereas Ξ*_L = z. For Q*_L ≈ z one needs the baseline weights w_i to encode the terminal expression values z_i; this is an additional parameter restriction that is not implied by enhancer co-binding or autoregulation. Since the energy function Eq. (6), the coarse-graining bound, the persistence-length estimate, and the blood progenitor ranking all depend on the symmetric form Eq. (5), this is a load-bearing gap. I recommend either proving the reduction under explicit conditions, testing Q ≈ Ξ directly on the data used for Ξ and on enhancer-to-gene assignment data, or showing that the main predictions are robust to finite Q − Ξ asymmetry.
  2. [Results, Eq. (8); Supplement B.4.2 and Supplement F] The persistence-length prediction β_L ≈ μ_L^{-2} for unequal-angle patterns is not justified by the stated bound. The bound |V_L − V| ≤ ¼ β μ_L^2 is small only when β μ_L^2 ≪ 1, but the equal-angle destabilization derived in Supplement F occurs at β_crit ≈ 2|L| μ^{-2}, i.e. at β μ_L^2 ≈ 2|L|. Thus, at the parameter regime where progenitor destabilization is claimed, the pointwise energy comparison is not small. The ranking of progenitor persistence lengths in Figure 6A therefore rests on an extrapolation beyond the regime in which the coarse-graining approximation is proved. A sharper bound or a separate stability analysis around the coarse-grained state at β ~ μ^{-2} is needed to support the central quantitative prediction.
minor comments (4)
  1. [Supplement B.3] The phrase 'Q*_L ≈ z since it is averaged of all Qi' is imprecise: the average of the rows Qi equals softmax(w_i), so the approximation holds only under a specific relation between w and z. State that relation explicitly.
  2. [Supplement F] There is a typo in the sentence 'all eigenvalues are negative if and only if Eq. 18 holds' — 'anf only if' should read 'if and only if'.
  3. [Figure 5 caption] The caption contains the typo 'correspondign' instead of 'corresponding'.
  4. [Eq. (7) and Supplement B.5] The main-text transformation Eq. (7) uses weights e^{w_i}, while the generalization in Supplement B.5 uses e^{w_i + β u_i}. Clarify in the main text which convention is used for the blood progenitor computations.

Circularity Check

1 steps flagged · score 6.0 of 10

The central reduction to symmetric gradient dynamics (Eq. 5) is an ansatz inherited from prior work; the claimed derivation via coarse-grained autoregulatory motifs requires an unstated assumption on the baseline weights w.

  1. ansatz smuggled in via citation [Supplement B.3, combined with Eq. 7 in the main text]
    "Within our modelling framework, such a motif corresponds to a set L such that for i∈L, we have that Ξ_i = z where z has large positive entries for all i∈L and zero entries for i∉L; and that Q_i has a large positive entry at index i and zero otherwise. Coarse-graining on L, which is associated with identical memory patterns, leaves the dynamics invariant; and its associated coarse-grained patterns are Ξ*_L = z and Q*_L ≈ z since it is averaged of all Q_i."

    By the paper's own Eq. 7, Q*_L = Σ_{i∈L} Q_i e^{w_i} / Σ e^{w_i}. Under B.3's stated motif Q_i = e_i (one-hot), Q*_L = softmax(w_i∈L), whereas Ξ*_L = z. The asserted equality Q*_L ≈ z 'since it is averaged of all Q_i' is therefore not a consequence of averaging; it holds only if the baseline weights w_i satisfy softmax(w_i) = z_i up to a constant. No such constraint is stated or biologically derived, and the model later treats w as a free signaling parameter (w = 0 in Fig. 4; feedback-fitted w in blood simulations). Since Eq. 5, the energy landscape Eq. 6, and all subsequent persistence-length/progenitor predictions require Q ≈ Ξ, they reduce to this unstated parametrization of w—an ansatz inherited from the authors' prior EnhancerNet paper [49]—rather than following from Eq.

full rationale

The paper's derivation chain from the general enhancer dynamics Eq. 4 to the symmetric gradient dynamics Eq. 5 is the load-bearing step for all subsequent results. In Supplement B.3, with the stated autoregulatory motif (Qi = e_i, Ξi = z), the coarse-grained matrix is Q*_L = softmax(w_i∈L) by Eq. 7, not z. The claim that Q*_L ≈ z 'since it is averaged of all Qi' is thus valid only if the weights w_i are chosen to encode the terminal pattern z_i. This is an unstated parameter assumption, not a consequence of enhancer competition or autoregulation, and the paper later treats w as freely adjustable. Therefore Eq. 5 is an ansatz imported from the authors' prior EnhancerNet framework [49] rather than a first-principles reduction of Eq. 4. This makes the central mechanistic claim partially circular: the hierarchy predictions are conditional on a symmetry that is effectively assumed by construction. That said, the coarse-graining transformation Eq. 7, its commutativity properties, and the energy calculations are internally derived mathematics; and the blood-progenitor test uses external Haemopedia terminal-expression data and externally curated lists of observed progenitors, so it is not circular in the statistical fitting sense. The validation is retrospective/rank-based rather than prospective, and the extrapolation from the O(βµ_L^2) bound to persistence length β_L ~ µ_L^{-2} is a rigor gap rather than a circular step. Overall: one load-bearing ansatz smuggled in via the authors' own prior work, giving partial circularity of the central claim.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central result rests on the EnhancerNet model (Eq. 4), the symmetric approximation (Eq. 5), and a set of simulation choices. The mathematical theorems in the Supplement are self-contained given these assumptions; the blood progenitor prediction adds no fitted parameters beyond the preprocessing of expression data, which is delegated to Karin (2024).

free parameters (6)
  • beta_max = 50
    Maximum value of beta in annealing simulations (Supplement G.1). Hand-chosen to ensure terminal attractors are stable; not fitted to data.
  • noise_amplitude_sigma = 0.05
    Amplitude of additive white noise in stochastic dynamics (Supplement Eq. 36). Chosen by hand to produce variability in annealing outcomes.
  • feedback_gains_Gp = 0.1, 0.01, 0.001
    Gains of the proportional feedback controller for w (Supplement G.1). Chosen to achieve balanced production; not calibrated to data.
  • initial_perturbation_zeta = Uniform[0, 0.1]
    Random initial-condition perturbation in annealing simulations (Supplement G.1).
  • erythropoietin_w_increase = 10% increase
    Hand-chosen perturbation to the erythrocyte w entry to simulate EPO effect (Results, Fig 6D).
  • annealing_parameters = t=20, n=1000
    Duration of annealing and number of cellular instances in simulations (Supplement G.1).
assumptions (6)
  • domain assumption The enhancer regulation model (Eq. 4) captures the dynamics of transcription-factor-driven gene expression through enhancer acetylation, Brd4 recruitment, and pause-release control.
    Derived in Appendix A from mechanistic steps, but these steps are idealizations, such as HATs limiting, quasi-steady-state acetylation, and Boltzmann redistribution of Brd4. Citations support each step, but the model is assumed.
  • domain assumption After coarse-graining autoregulatory motifs, the enhancer-binding matrix Xi equals the expression-contribution matrix Q, giving the symmetric dynamics of Eq. 5.
    This is the key reduction from Eq. 4 to Eq. 5; it is stated without a quantitative error bound for realistic networks and is the weakest assumption in the paper.
  • standard math The rows of Xi (terminal expression programs) are of unit magnitude in the main analysis; unequal magnitudes are handled only in the Appendix.
    The coarse-graining bounds and beta_crit derivation assume unit-norm patterns; the generalization is sketched but not used in the blood predictions.
  • domain assumption Cell-type expression profiles from Haemopedia can serve directly as the memory patterns Xi, i.e., each terminal cell type is a stable attractor of the dynamics.
    Used to initialize Xi; this presumes that observed expression is at an attractor, and the mapping from expression data to Xi involves preprocessing choices defined in Karin (2024).
  • domain assumption Differentiation and annealing are implemented as a decrease followed by an increase in beta, with an additive Gaussian noise term in the dynamics.
    The annealing protocol and noise model are simulation choices (Supplement G.1); the noise is not derived from biological stochasticity.
  • domain assumption The feedback of w through proportional error control approximates physiological regulation of blood cell production.
    Used in the balanced-differentiation simulations; a modeling choice without direct experimental calibration.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hierarchical cell identities emerge from animal gene regulatory mechanisms." pith.science (2026). https://pith.science/paper/M7P56L4G

@misc{pith2026241211336,
  author       = {Pith},
  title        = {Pith review of: Hierarchical cell identities emerge from animal gene regulatory mechanisms},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M7P56L4G}},
  note         = {Machine review of arXiv:2412.11336}
}
read the original abstract

The hierarchical organisation of cell identity is a fundamental feature of animal development with rich and well-characterized experimental phenomenology, yet the mechanisms driving its emergence remain unknown. The regulation of cell identity genes relies on a distinct mechanism involving higher-order interactions of transcription factors on distant regulatory regions called enhancers. These interactions are mediated by epigenetic regulators that are broadly shared between enhancers. Through the development of a new and predictive mathematical theory on the effects of epigenetic regulator activity on gene network dynamics, we demonstrate that hierarchical identities are essential emergent properties of animal-specific gene regulatory mechanisms. Hierarchical identities arise from the interplay between enhancer competition for epigenetic readers and cooperation through activation of shared transcriptional programs. We show that epigenetic regulatory mechanisms provide the network with self-similar properties that enable multilineage priming and signal-dependent control of progenitor states. The stabilisation of progenitor states is predicted to be controlled by the balance in activities between epigenetic writers and erasers. Our model quantitatively predicts lineage relationships, reconstructs all known blood progenitor states from terminal states, and explains mechanisms of cell identity dysregulation in cancer and the general differentiation effects of histone deacetylase inhibition. We identify non-specific modulation of enhancer competition as a central regulatory axis, with implications for developmental biology, cancer, and differentiation therapy.

Figures

Figures reproduced from arXiv: 2412.11336 by the authors.

Figure 1
Figure 1. Example of application of transformation for patterns of arbitrary magnitude. [PITH_FULL_IMAGE:figures/full_fig_p038_1.png] view at source ↗
Figure 2
Figure 2. An example of how modulating w1 moves the separatrix between two patterns In this example we have 2 patterns, represented by black dots. Ξ1 is the pattern on the left. The red dashed line represents the analytic estimate for the separatrix. In this simulation β = 7, w = (w1, 0). E.2.1 Effect of a far away pattern on the separatrix In this section, we will show that an addition of a pattern that is far away from the … view at source ↗
Figure 3
Figure 3. An example of how modulating w1 moves the separatrix of the coarse-grained pattern In this example we have 3 patterns, represented by black dots, with a coarse-grained pattern (gray dot), stemming from Ξ1 and Ξ2. Ξ1 is the upper right pattern. The red dashed line represents the analytic estimate for the separatrix. In this simulation β = 7, w = (w1, 1, 0). As seen w1 has little effect on the position of the separatr… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 38 canonical work pages

  1. [1]

    author author J. J. \ Hopfield ,\ title title Neural networks and physical systems with emergent collective computational abilities. ,\ @noop journal journal Proceedings of the national academy of sciences \ volume 79 ,\ pages 2554 ( year 1982 ) NoStop

  2. [2]

    author author D. J. \ Amit , author H. Gutfreund ,\ and\ author H. Sompolinsky ,\ title title Storing infinite numbers of patterns in a spin-glass model of neural networks ,\ @noop journal journal Physical Review Letters \ volume 55 ,\ pages 1530 ( year 1985 ) NoStop

  3. [3]

    Krotov \ and\ author J

    author author D. Krotov \ and\ author J. J. \ Hopfield ,\ title title Dense associative memory for pattern recognition ,\ @noop journal journal Advances in neural information processing systems \ volume 29 ( year 2016 ) NoStop

  4. [4]

    Demircigil , author J

    author author M. Demircigil , author J. Heusel , author M. L \"o we , author S. Upgang ,\ and\ author F. Vermet ,\ title title On a model of associative memory with huge storage capacity ,\ @noop journal journal Journal of Statistical Physics \ volume 168 ,\ pages 288 ( year 2017 ) NoStop

  5. [5]

    Krotov \ and\ author J

    author author D. Krotov \ and\ author J. Hopfield ,\ title title Large associative memory problem in neurobiology and machine learning ,\ @noop journal journal arXiv preprint arXiv:2008.06996 \ ( year 2020 ) NoStop

  6. [6]

    Ramsauer , author B

    author author H. Ramsauer , author B. Sch \"a fl , author J. Lehner , author P. Seidl , author M. Widrich , author T. Adler , author L. Gruber , author M. Holzleitner , author M. Pavlovi \'c , author G. K. \ Sandve , et al. ,\ title title Hopfield networks is all you need ,\ @noop journal journal arXiv preprint arXiv:2008.02217 \ ( year 2020 ) NoStop

  7. [7]

    Lucibello \ and\ author M

    author author C. Lucibello \ and\ author M. M \'e zard ,\ title title Exponential capacity of dense associative memories ,\ @noop journal journal Physical Review Letters \ volume 132 ,\ pages 077301 ( year 2024 ) NoStop

  8. [8]

    author author A. H. \ Lang , author H. Li , author J. J. \ Collins ,\ and\ author P. Mehta ,\ title title Epigenetic landscapes explain partially reprogrammed cells and identify key reprogramming genes ,\ @noop journal journal PLoS computational biology \ volume 10 ,\ pages e1003734 ( year 2014 ) NoStop

Show all 42 references
  1. [9]

    author author N. E. \ Boukacem , author A. Leary , author R. Th \'e riault , author F. Gottlieb , author M. Mani ,\ and\ author P. Fran c ois ,\ title title Waddington landscape for prototype learning in generalized hopfield networks ,\ @noop journal journal Physical Review Re...

  2. [10]

    author author S. T. \ Pusuluri , author A. H. \ Lang , author P. Mehta ,\ and\ author H. E. \ Castillo ,\ title title Cellular reprogramming dynamics follow a simple 1d reaction coordinate ,\ @noop journal journal Physical Biology \ volume 15 ,\ pages 016001 ( year 2017 ) NoStop

  3. [11]

    author author O. Karin ,\ title title Enhancernet: A predictive model of cell identity dynamics through enhancer selection ,\ @noop journal journal Development \ volume 151 ,\ pages dev202997 ( year 2024 ) NoStop

  4. [12]

    Narita , author S

    author author T. Narita , author S. Ito , author Y. Higashijima , author W. K. \ Chu , author K. Neumann , author J. Walter , author S. Satpathy , author T. Liebner , author W. B. \ Hamilton , author E. Maskey , et al. ,\ title title Enhancers are activated by p300/cbp activit...

  5. [13]

    author author J. J. \ Ferrie , author J. P. \ Karr , author T. G. \ Graham , author G. M. \ Dailey , author G. Zhang , author R. Tjian ,\ and\ author X. Darzacq ,\ title title p300 is an obligate integrator of combinatorial transcription factor inputs ,\ @noop journal journal ...

  6. [14]

    author author L. Ambrogioni ,\ title title In search of dispersed memories: Generative diffusion models are associative memory networks ,\ @noop journal journal arXiv preprint arXiv:2309.17290 \ ( year 2023 ) NoStop

  7. [15]

    Spens \ and\ author N

    author author E. Spens \ and\ author N. Burgess ,\ title title A generative model of memory construction and consolidation ,\ @noop journal journal Nature Human Behaviour \ volume 8 ,\ pages 526 ( year 2024 ) NoStop

  8. [16]

    Pham , author G

    author author B. Pham , author G. Raya , author M. Negri , author M. J. \ Zaki , author L. Ambrogioni ,\ and\ author D. Krotov ,\ title title Memorization to generalization: The emergence of diffusion models from associative memory ,\ in\ @noop booktitle NeurIPS 2024 Workshop ...

  9. [17]

    Hnisz , author B

    author author D. Hnisz , author B. J. \ Abraham , author T. I. \ Lee , author A. Lau , author V. Saint-Andr \'e , author A. A. \ Sigova , author H. A. \ Hoke ,\ and\ author R. A. \ Young ,\ title title Super-enhancers in the control of cell identity and disease ,\ @noop journa...

  10. [18]

    Saint-Andr \'e , author A

    author author V. Saint-Andr \'e , author A. J. \ Federation , author C. Y. \ Lin , author B. J. \ Abraham , author J. Reddy , author T. I. \ Lee , author J. E. \ Bradner ,\ and\ author R. A. \ Young ,\ title title Models of human core transcriptional regulatory circuitries ,\ ...

  11. [19]

    Stanoev \ and\ author A

    author author A. Stanoev \ and\ author A. Koseska ,\ title title Robust cell identity specifications through transitions in the collective state of growing developmental systems ,\ @noop journal journal Current Opinion in Systems Biology \ volume 31 ,\ pages 100437 ( year 2022...

  12. [20]

    author author B. D. \ Simons \ and\ author O. Karin ,\ title title Tuning of plasma cell lifespan by competition explains the longevity and heterogeneity of antibody persistence ,\ @noop journal journal Immunity \ volume 57 ,\ pages 600 ( year 2024 ) NoStop

  13. [21]

    Grover , author E

    author author A. Grover , author E. Mancini , author S. Moore , author A. J. \ Mead , author D. Atkinson , author K. D. \ Rasmussen , author D. O’Carroll , author S. E. W. \ Jacobsen ,\ and\ author C. Nerlov ,\ title title Erythropoietin guides multipotent hematopoietic progen...

  14. [22]

    author author A. S. \ Eisele , author J. Cosgrove , author A. Magniez , author E. Tubeuf , author S. T. \ Bento , author C. Conrad , author F. Cayrac , author T. Tak , author A.-M. \ Lyne , author J. Urbanus , et al. ,\ title title Erythropoietin directly remodels the clonal c...

  15. [23]

    Soldatov , author M

    author author R. Soldatov , author M. Kaucka , author M. E. \ Kastriti , author J. Petersen , author T. Chontorotzea , author L. Englmaier , author N. Akkuratova , author Y. Yang , author M. H \"a ring , author V. Dyachuk , et al. ,\ title title Spatiotemporal structure of cel...

  16. [24]

    author author M. X. \ Moreau , author Y. Saillour , author A. W. \ Cwetsch , author A. Pierani ,\ and\ author F. Causeret ,\ title title Single-cell transcriptomics of the early developing mouse cerebral cortex disentangle the spatial and temporal components of neuronal fate a...

  17. [25]

    Nusser , author Sagar , author J

    author author A. Nusser , author Sagar , author J. B. \ Swann , author B. Krauth , author D. Diekhoff , author L. Calderon , author C. Happe , author D. Gruen ,\ and\ author T. Boehm ,\ title title Developmental dynamics of two bipotent thymic epithelial progenitor types ,\ @n...

  18. [26]

    Choi , author T

    author author J. Choi , author T. M. \ Baldwin , author M. Wong , author J. E. \ Bolden , author K. A. \ Fairfax , author E. C. \ Lucas , author R. Cole , author C. Biben , author C. Morgan , author K. A. \ Ramsay , et al. ,\ title title Haemopedia rna-seq: a database of gene ...

  19. [27]

    Ceredig , author A

    author author R. Ceredig , author A. G. \ Rolink ,\ and\ author G. Brown ,\ title title Models of haematopoiesis: seeing the wood for the trees ,\ @noop journal journal Nature Reviews Immunology \ volume 9 ,\ pages 293 ( year 2009 ) NoStop

  20. [28]

    Laurenti \ and\ author B

    author author E. Laurenti \ and\ author B. G \"o ttgens ,\ title title From haematopoietic stem cells to complex differentiation landscapes ,\ @noop journal journal Nature \ volume 553 ,\ pages 418 ( year 2018 ) NoStop

  21. [29]

    Brown \ and\ author I

    author author G. Brown \ and\ author I. Sanchez-Garcia ,\ title title Is lineage decision-making restricted during tumoral reprograming of haematopoietic stem cells? ,\ @noop journal journal Oncotarget \ volume 6 ,\ pages 43326 ( year 2015 ) NoStop

  22. [30]

    Wanet , author M

    author author A. Wanet , author M. A. \ Bassal , author S. B. \ Patel , author F. Marchi , author S. A. \ Mariani , author N. Ahmed , author H. Zhang , author M. Borchiellini , author S. Chen , author J. Zhang , et al. ,\ title title E-cadherin is regulated by gata-2 and marks...

  23. [31]

    Park \ and\ author S.-J

    author author J. Park \ and\ author S.-J. \ Kang ,\ title title The ontogenesis and heterogeneity of basophils ,\ @noop journal journal Discovery Immunology \ volume 3 ,\ pages kyae003 ( year 2024 ) NoStop

  24. [32]

    Metcalf , author S

    author author D. Metcalf , author S. Mifsud ,\ and\ author L. Di Rago ,\ title title Stem cell factor can stimulate the formation of eosinophils by two types of murine eosinophil progenitor cells ,\ @noop journal journal Stem Cells \ volume 20 ,\ pages 460 ( year 2002 ) NoStop

  25. [33]

    Kim , author S

    author author K. Kim , author S. M. \ Hwang , author S. M. \ Kim , author S. W. \ Park , author Y. Jung ,\ and\ author I. Y. \ Chung ,\ title title Terminally differentiating eosinophils express neutrophil primary granule proteins as well as eosinophil-specific granule protein...

  26. [34]

    Kwok , author E

    author author I. Kwok , author E. Becht , author Y. Xia , author M. Ng , author Y. C. \ Teh , author L. Tan , author M. Evrard , author J. L. \ Li , author H. T. \ Tran , author Y. Tan , et al. ,\ title title Combinatorial single-cell analyses of granulocyte-monocyte progenito...

  27. [35]

    author author C. H. \ Waddington ,\ @noop title The strategy of the genes \ ( publisher Routledge ,\ year 2014 ) NoStop

  28. [36]

    author author R. A. \ Nimmo , author G. E. \ May ,\ and\ author T. Enver ,\ title title Primed and ready: understanding lineage commitment through single cell analysis ,\ @noop journal journal Trends in cell biology \ volume 25 ,\ pages 459 ( year 2015 ) NoStop

  29. [37]

    author author A. G. \ Erickson , author P. Kameneva ,\ and\ author I. Adameyko ,\ title title The transcriptional portraits of the neural crest at the individual cell level ,\ in\ @noop booktitle Seminars in cell & developmental biology ,\ Vol.\ volume 138 \ ( organization Els...

  30. [38]

    Enhancernet: A predictive model of cell identity dynamics through enhancer selection

    Omer Karin. Enhancernet: A predictive model of cell identity dynamics through enhancer selection. Development , 151(19):dev202997, 2024

  31. [39]

    Enhancers are activated by p300/cbp activity-dependent pic assembly, rnapii recruitment, and pause release

    Takeo Narita, Shinsuke Ito, Yoshiki Higashijima, Wai Kit Chu, Katrin Neumann, Jonas Walter, Shankha Satpathy, Tim Liebner, William B Hamilton, Elina Maskey, et al. Enhancers are activated by p300/cbp activity-dependent pic assembly, rnapii recruitment, and pause release. Molec...

  32. [40]

    Super-enhancers in the control of cell identity and disease

    Denes Hnisz, Brian J Abraham, Tong Ihn Lee, Ashley Lau, Violaine Saint-Andr \'e , Alla A Sigova, Heather A Hoke, and Richard A Young. Super-enhancers in the control of cell identity and disease. Cell , 155(4):934--947, 2013

  33. [41]

    Models of human core transcriptional regulatory circuitries

    Violaine Saint-Andr \'e , Alexander J Federation, Charles Y Lin, Brian J Abraham, Jessica Reddy, Tong Ihn Lee, James E Bradner, and Richard A Young. Models of human core transcriptional regulatory circuitries. Genome research , 26(3):385--396, 2016

  34. [42]

    Haemopedia rna-seq: a database of gene expression during haematopoiesis in mice and humans

    Jarny Choi, Tracey M Baldwin, Mae Wong, Jessica E Bolden, Kirsten A Fairfax, Erin C Lucas, Rebecca Cole, Christine Biben, Clare Morgan, Kerry A Ramsay, et al. Haemopedia rna-seq: a database of gene expression during haematopoiesis in mice and humans. Nucleic acids research , 4...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.