Pith. sign in

REVIEW 5 major objections 6 minor 37 references

Active learning for efficient discovery of optimal gene combinations in the combinatorial perturbation space

T0 review · 5 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read NAIAD claims the combinatorial perturbation space becomes tractable when a model is conditioned on measured single-gene effects and uses adaptive gene embeddings, and that a maximum-predicted-effect acquisition strategy finds most of the…

desk verdict NAIAD is a useful active learning framework for combinatorial CRISPR screens, but the 40% claim is overstated and the MPE discovery result lacks a linear-surrogate control. read the letter →

arxiv 2411.12010 v2 pith:72ROTDRC submitted 2024-11-18 q-bio.GN cs.LG

Jason Qin , Hans-Hermann Wessels , Carlos Fernandez-Granda , Yuhan Hao This is my paper · ORCID
classification q-bio.GNcs.LG
keywords activelearningCRISPRcombinatorialscreensgene-geneinteractionsgeneembeddingsmaximumpredictedeffectsamplingsmall-samplecellviabilityacquisitionfunction
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

NAIAD is an active learning framework for the combinatorial perturbation space—the set of all two-gene knockouts or activations that could push a cell into a desired state, which is astronomically large (~200 million pairs for the human genome). The paper's central claim is that a model which takes measured single-gene effects as inputs and uses gene embeddings whose dimensionality grows with the training set can predict unseen gene-pair phenotypes accurately from very small samples, beating both simple linear models and deep baselines. It further claims that selecting the next experiments by maximum predicted effect (MPE), rather than by uncertainty or a mixed bound, accelerates discovery: after four rounds the MPE strategy recovers over twice as many of the top-200 strongest pairs as uniform sampling and nearly 1.5 times as many as UCB. If correct, the framework lets labs design CRISPR libraries that converge on potent combinations in a handful of experiment–model iterations instead of exhaustive screens.

What carries the argument

The model's core object is a decomposition of the pairwise phenotype into two terms: an over-parameterized single-gene effect term $\phi([Y_i,Y_j]W_1)A_1^T$ that conditions on measured single-gene effects $Y_i, Y_j$, and an interaction term $f(\phi(W_2X^i_{\text{gene}}), \phi(W_2X^j_{\text{gene}}))A_2^T$ built from adaptive gene embeddings $X_{\text{gene}}$ with a permutation-invariant combination (sum) and a nonlinear encoder $f$. The embedding dimension starts small and grows on a predetermined schedule tied to the average gene-occurrence frequency, which lets the model behave like a linear model when data are scarce and like a deep network when data accumulate. The acquisition side relies on an ensemble of NAIAD models whose prediction variance estimates uncertainty, and the recommended strategy is Maximum Predicted Effects (MPE) sampling, which selects the gene pairs with the largest absolute predicted effects for the next experimental round.

What would settle it

Run a prospective four-round CRISPR screen where NAIAD's MPE sampling and UCB sampling each choose 100 new pairs per round, then count how many of the top-200 ground-truth pairs each strategy finds; if MPE does not recover more than UCB, the paper's central active-learning claim fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the combinatorial perturbation prediction problem has a favorable small-sample structure when it is decomposed correctly. NAIAD writes the phenotype of a gene pair as the sum of an over-parameterized transformation of the two single-gene effects and a nonlinear interaction term computed from learnable gene embeddings, with the embedding dimension deliberately scheduled to grow with the average number of times each gene has been observed. Across four bulk CRISPR datasets totaling over 350,000 measured interactions, NAIAD trained on small samples achieves the lowest RMSE, highest correlation, and highest true-positive rate among five models, and its advantage is largest when each gene appears only about four times in the training set. The paper also shows that the MPE acquisition function, despite producing worse global MSE because it skews training data toward strong pairs, is the most efficient at finding the strongest gene pairs, which is the stated objective for therapeutic discovery.

Load-bearing premise

NAIAD assumes the single-gene perturbation effects for every gene in a pair are known and correct; if those measurements are noisy, biased, or missing, the model's additive foundation is corrupted and its advantage could vanish.

Editorial extensions

If this is right

  • At roughly four observations per gene, NAIAD's test RMSE is over 40% better than the second-best model across the four datasets.
  • The MPE acquisition function identifies about 150 of the top-200 strongest gene pairs by round four in three of the four screens, versus 66–92 for uniform sampling.
  • Because genome-wide coverage at 20 observations per gene would require roughly 200,000 initial combinations, cutting the needed observations per gene is what makes genome-scale screens feasible.
  • Adaptive embeddings give a single model that remains competitive in both the small-sample and data-rich regimes, removing the usual need to switch from linear to deep models as data grow.

Reading between the lines

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

  • Because NAIAD's additive term is conditioned on measured single-gene effects, its advantage is likely to shrink as single-gene measurements get noisier; a sensitivity analysis sweeping single-gene noise would test this directly.
  • The top-k discovery metric used to evaluate acquisition functions is the right objective for therapeutic library design, and its widespread adoption would make comparisons between active learning methods more meaningful than global MSE.
  • The embedding-growth schedule is a general trick for sparse combinatorial search and could transfer to drug combinations or multi-target screens, though the schedule constants would need re-tuning.
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

5 major / 6 minor

Summary. The paper introduces NAIAD, an active learning framework for predicting the phenotypic effects of two-gene perturbations in combinatorial CRISPR screens. The model architecture combines an overparameterized single-gene effect term (a high-dimensional projection of the measured single-gene effects Yi and Yj) with learnable gene embeddings whose dimension grows according to a schedule based on training-set size. For active learning, the paper proposes maximum predicted effect (MPE) sampling, which selects the gene pairs with the largest predicted effects for the next experimental round. The authors evaluate NAIAD on four bulk combinatorial perturbation datasets (Norman, Simpson, Horlbeck K562, Horlbeck Jurkat) and compare against linear regression, MLP, GEARS, and RECOVER, reporting that NAIAD outperforms the baselines in small-sample settings, with a claimed 'up to 40%' relative improvement over the second-best model. They further report that MPE sampling uncovers more than twice as many of the top-200 strongest pairs as uniform sampling and nearly 1.5 times as many as UCB by round four of the simulated active learning loop.

Significance. If the empirical claims are correct, NAIAD would be a practically useful tool for designing combinatorial CRISPR libraries in the small-data regime, where active learning is most needed. The adaptive embedding dimension is a sensible and interesting mechanism for trading off model capacity against data availability, and the paper evaluates on unusually large public datasets with held-out test sets, which is a strength. The active learning simulation protocol is also a reasonable proxy for real iterative experimentation. However, the headline quantitative claims are not fully supported by the reported results, and the benchmark and control choices weaken the causal link between the architecture and the reported discovery gains. Specifically, the 40% improvement figure is not reproducible from Table 1, the GEARS baseline is adapted in a way that likely handicaps it, and the MPE acquisition advantage is not compared against a simple linear surrogate, leaving open the possibility that additive single-gene effects, rather than the learned interaction embeddings, drive the discovery performance.

major comments (5)
  1. [Abstract; Section 5.2, Table 1] The Abstract and Section 5.2 state that NAIAD outperforms the second-best model by 'over 40%' on average across the four datasets based on Table 1. This is not supported by the reported RMSE values. At gene frequency 4, the relative improvement over the second-best model (the linear model in all four cases) is approximately 18% for Norman, 33% for Simpson, 5% for Horlbeck K562, and 23% for Horlbeck Jurkat, an average of about 20%. Please correct the claim or clarify the exact calculation, since the current phrasing is materially misleading.
  2. [Appendix B (GEARS benchmark)] The description of the GEARS baseline states that for bulk datasets the authors 'generated synthetic Perturb-seq datasets of normalized gene expression matrices, with a separate Gaussian N(0,1) used for sampling the expression of each gene.' GEARS is designed to take single-cell transcriptomic profiles as input; feeding it random Gaussian features is not a meaningful or fair benchmark and most likely explains GEARS's poor performance. To support the claim that NAIAD outperforms GEARS, the comparison must use a reasonable input representation (for example, pseudobulk expression derived from the same screen or a compatible single-cell dataset), or the GEARS baseline should be removed.
  3. [Section 5.3, Table 2, Figure 4] The MPE acquisition function is evaluated only with the NAIAD surrogate. The paper does not include a control where MPE ranks candidate pairs by a linear model (e.g., by Yi + Yj or by a linear regression fit on single-gene effects). Given that Appendix Figure 5 shows NAIAD's single-gene component strongly correlates with linear model predictions, the reported MPE advantage over uniform and UCB sampling could be fully explained by additive single-gene ranking rather than by NAIAD's learned interactions or adaptive embeddings. Without such a control, the claim that NAIAD's architecture is responsible for the discovery efficiency is not established.
  4. [Section 3.2, Appendix F.1, Table 5] The adaptive embedding schedule is described as a general principle that 'hold[s] across all datasets,' but Table 5 gives different embedding dimensions for the same gene occurrence frequency in different datasets. For example, at gene frequency 4, Norman and Simpson use embedding dimension 4 while Horlbeck K562 uses 16; at frequency 10, Norman and Simpson use 16 while Horlbeck uses 32. This indicates that the schedule is dataset-specific rather than a single shared rule. The authors should either present one schedule valid across datasets or explicitly state that the per-dataset values are tuned hyperparameters, and explain how they were chosen without using test information.
  5. [Section 3.1, Section 6 (Discussion)] The model uses measured single-gene effects Yi as inputs, and the Discussion acknowledges this assumption. However, the paper provides no sensitivity analysis for the impact of noise, bias, or missing values in Yi on either predictive accuracy or the MPE-based discovery results. Since the additive component and the MPE ranking depend directly on these values, a robustness study (e.g., adding Gaussian noise to Yi, or removing a subset of single-gene measurements) is needed to support the practical applicability claimed in the paper.
minor comments (6)
  1. [Table 2, Table 3, Figure 4] Table 2 and Table 3 report point values without standard errors, while Figure 4 for the same experiment reports SE across three cross-fold replicates. Please clarify whether the tables report a representative fold or an average, and consider adding uncertainty estimates to the tables.
  2. [Appendix D] The definition of 'Fraction Discovered' concatenates model predictions on unseen data with measured values from seen data and then takes the top P; this is a nonstandard mixture. The main text should explain this more clearly, as it affects the interpretation of the 'top 200 discovered' numbers in Table 2.
  3. [Figures 2, 3, 5, 7] The figures report log(MSE) without specifying the logarithm base (natural or base 10). Please state this in the captions or methods.
  4. [Section 3.1, model equation] The dimensions of the matrices W1, W2, A1, and A2 are not explicitly defined. Please specify all tensor dimensions so that the architecture is unambiguous.
  5. [Figure 2 caption] The caption refers to '4429 training combinations' for the Norman dataset, while the dataset description in Section 4.2 and Appendix A states 6,328 unique combinations. Please clarify the number and the split used.
  6. [Section 5.3 and Table 2] The text in Section 5.3 says 'MPE sampling ranked the top 200 gene pairs' but Table 2 reports 'Number of top 200 gene combinations correctly discovered at round 4.' Please ensure the terminology is consistent and that the reader understands whether the table counts pairs that were measured, predicted, or either.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: NAIAD's claims rest on held-out measured outcomes; single-gene effects are inputs, and no fitted parameter is relabeled as a prediction.

full rationale

The paper's derivation chain is self-contained against external benchmarks. NAIAD is trained on a training split and evaluated with log(MSE), RMSE, Pearson correlation, and TPR on a disjoint 30% held-out test set (Section 4.2, Section 5.2, Appendix D); no target quantity is defined in terms of the model's own outputs. Single-gene perturbation effects Yi are measured inputs to Eq. (1), not fitted parameters or predicted variables. The MPE acquisition function ranks candidate pairs by the model's predicted effect and selects the top ones to measure; the measured outcomes are then counted in the 'Fraction Discovered' metric (Appendix D). This is the standard active-learning objective, not circularity: the selected pairs are counted only because they are measured, and the value of the selection depends on genuine predictive accuracy, which is independently established in Section 5.2. The adaptive embedding schedule (Table 5) is a capacity-control choice, not a definition of the target. Section 5.1's observation that the compressed single-gene component correlates with the linear model is an empirical characterization, and the absence of an MPE-with-linear-surrogate baseline (the skeptic's concern) is a missing control that bears on attribution of the discovery gain to the interaction model; it is a correctness risk, not a circular step. The Discussion's limitation that 'our model assumes information of each gene's individual effect on the phenotype' is an acknowledged scope restriction, and it does not make any prediction equivalent to its inputs. No load-bearing self-citations or imported uniqueness theorems appear; RECOVER, GEARS, and other baselines are external methods. I therefore find no step that reduces, by construction or by self-citation, to the paper's inputs.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

NAIAD introduces no new biological entities. The main assumptions are measurement quality and comparability of single-gene and pair effects. The adaptive embedding schedule is the principal hand-fitted element.

free parameters (3)
  • Adaptive embedding dimension schedule = Table 5: dim 2 to 128 depending on dataset and average gene occurrence
    The embedding dimension p is chosen by hand per dataset based on estimated gene frequency, not derived from theory. This affects model capacity and is a free parameter.
  • Single-gene projection dimension = 64 for Norman, 256 for Simpson and Horlbeck
    The dimension of the overparameterized single-gene effect encoder W1 is a hyperparameter selected via grid search (Appendix F).
  • Training hyperparameters (learning rate, batch size, epochs) = lr=1e-2, batch=1024, epochs 200-500
    Selected over ranges in Appendix F; these are standard hyperparameters but still fitted to the datasets.
assumptions (3)
  • domain assumption Single-gene perturbation effects Yi are accurately measured for every gene in the screen.
    The model uses Yi and Yj as inputs to the additive term. The Discussion admits that for unseen genes the current model cannot predict combinations.
  • domain assumption Cell viability measurements from different bulk screens are comparable and complete.
    The evaluation pools four datasets from different labs and protocols; no batch correction or noise model is applied.
  • domain assumption Held-out test data approximates future experimental performance under active learning selection.
    The active learning loop is simulated by sampling from the same screen, assuming no batch effects or time-varying measurement error.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Active learning for efficient discovery of optimal gene combinations in the combinatorial perturbation space." pith.science (2026). https://pith.science/paper/72ROTDRC

@misc{pith2026241112010,
  author       = {Pith},
  title        = {Pith review of: Active learning for efficient discovery of optimal gene combinations in the combinatorial perturbation space},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/72ROTDRC}},
  note         = {Machine review of arXiv:2411.12010}
}
read the original abstract

The advancement of novel combinatorial CRISPR screening technologies enables the identification of synergistic gene combinations on a large scale. This is crucial for developing novel and effective combination therapies, but the combinatorial space makes exhaustive experimentation infeasible. We introduce NAIAD, an active learning framework that efficiently discovers optimal gene pairs capable of driving cells toward desired cellular phenotypes. NAIAD leverages single-gene perturbation effects and adaptive gene embeddings that scale with the training data size, mitigating overfitting in small-sample learning while capturing complex gene interactions as more data is collected. Evaluated on four CRISPR combinatorial perturbation datasets totaling over 350,000 genetic interactions, NAIAD, trained on small datasets, outperforms existing models by up to 40\% relative to the second-best. NAIAD's recommendation system prioritizes gene pairs with the maximum predicted effects, resulting in the highest marginal gain in each AI-experiment round and accelerating discovery with fewer CRISPR experimental iterations. Our NAIAD framework (https://github.com/NeptuneBio/NAIAD) improves the identification of novel, effective gene combinations, enabling more efficient CRISPR library design and offering promising applications in genomics research and therapeutic development.

Figures

Figures reproduced from arXiv: 2411.12010 by the authors.

Figure 1
Figure 1. Illustration of active learning framework in CRISPR combinatorial perturbation (A) and our [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Performance on test data of different gene embedding settings in NAIAD using the [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Benchmark analysis comparing the NAIAD model with GEARS and RECOVER models, [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparison of different acquisition functions evaluated by top [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Evaluation of compressed single-gene effects and gene embeddings shows a strong [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Benchmark analysis comparing the NAIAD model with GEARS and RECOVER models, [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Comparison of different acquisition functions evaluated by MSE across four iteration [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Comparison of training sample distribution changes across different iterations using four [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 27 canonical work pages

  1. [1]

    Deep Learning-Based Predictions of Gene Perturbation Effects Do Not yet Outperform Simple Linear Methods

    Ahlmann-Eltze, Constantin, Wolfgang Huber, and Simon Anders. 2024. “Deep Learning-Based Predictions of Gene Perturbation Effects Do Not yet Outperform Simple Linear Methods.” bioRxiv. https://doi. org/10.1101/2024.09.16.613342

    doi:10.1101/2024.09.16.613342 2024

  2. [2]

    Learning to Make Decisions via Submodular Regularization

    Alieva, Ayya, Aiden Aceves, Jialin Song, Stephen Mayo, Yisong Yue, and Yuxin Chen. 2020. “Learning to Make Decisions via Submodular Regularization.” InInternational Conference on Learning Representations. https://openreview.net/pdf?id=ac288vnG_7U

    work page 2020

  3. [3]

    Combinatorial Drug Therapy for Cancer in the Post-Genomic Era

    Al-Lazikani, Bissan, Udai Banerji, and Paul Workman. 2012. “Combinatorial Drug Therapy for Cancer in the Post-Genomic Era.” Nature Biotechnology, 30 (7): 679–92

    work page 2012

  4. [4]

    Modelling Cellular Perturbations with the Sparse Additive Mechanism Shift Variational Autoencoder

    Bereket, Michael, and Theofanis Karaletsos. 2023. “Modelling Cellular Perturbations with the Sparse Additive Mechanism Shift Variational Autoencoder.”http://arxiv.org/abs/2311.02794

    work page Pith review arXiv 2023

  5. [5]

    RECOVER Identifies Synergistic Drug Combinations in Vitro through Sequential Model Optimization

    Bertin, P., J. Rector-Brooks, D. Sharma, T. Gaudelet, A. Anighoro, T. Gross, F. Martínez-Peña, et al. 2023. “RECOVER Identifies Synergistic Drug Combinations in Vitro through Sequential Model Optimization.” Cell Reports Methods, 3 (10). https://doi.org/10.1016/j.crmeth.2023.100599

    arXiv 2023

  6. [6]

    On Initial Pools for Deep Active Learning

    Chandra, Akshay L., Sai Vikas Desai, Chaitanya Devaguptapu, and Vineeth N. Balasubramanian. 2020. “On Initial Pools for Deep Active Learning.” http://arxiv.org/abs/2011.14696

    work page Pith review arXiv 2020

  7. [7]

    scGPT: Toward Building a Foundation Model for Single-Cell Multi-Omics Using Generative AI

    Cui, Haotian, Chloe Wang, Hassaan Maan, Kuan Pang, Fengning Luo, Nan Duan, and Bo Wang. 2024. “scGPT: Toward Building a Foundation Model for Single-Cell Multi-Omics Using Generative AI.”Nature Methods, 21 (8): 1470–80

    work page 2024

  8. [8]

    Active Machine Learning Helps Drug Hunters Tackle Biology

    Eisenstein, Michael. 2020. “Active Machine Learning Helps Drug Hunters Tackle Biology.” Nature Biotechnology, 38 (5): 512–14

    work page 2020

Show all 37 references
  1. [9]

    A Tutorial on Bayesian Optimization

    Frazier, Peter I. 2018. “A Tutorial on Bayesian Optimization.” http://arxiv.org/abs/1807.02811

    2018 arXiv

  2. [10]

    Season Combinatorial Intervention Predictions with Salt & Peper

    Gaudelet, Thomas, Alice Del Vecchio, Eli M. Carrami, Juliana Cudini, Chantriolnt-Andreas Kapourani, Caroline Uhler, and Lindsay Edwards. 2024. “Season Combinatorial Intervention Predictions with Salt & Peper.” http://arxiv.org/abs/2404.16907

    2024 arXiv

  3. [11]

    Systematically Characterizing the Roles of E3-Ligase Family Members in Inflammatory Responses with Massively Parallel Perturb-Seq

    Geiger-Schuller, Kathryn, Basak Eraslan, Olena Kuksenko, Kushal K. Dey, Karthik A. Jagadeesh, Pratiksha I. Thakore, Ozge Karayel, et al. 2023. “Systematically Characterizing the Roles of E3-Ligase Family Members in Inflammatory Responses with Massively Parallel Perturb-Seq.” b...

    2023 doi

  4. [12]

    Adaptive Submodularity: Theory and Applications in Active Learning and Stochastic Optimization

    Golovin, Daniel, and Andreas Krause. 2010. “Adaptive Submodularity: Theory and Applications in Active Learning and Stochastic Optimization.” http://arxiv.org/abs/1003.3967

    2010 arXiv

  5. [13]

    Oral Nirmatrelvir for High-Risk, Nonhospitalized Adults with Covid-19

    Hammond, Jennifer, Heidi Leister-Tebbe, Annie Gardner, Paula Abreu, Weihang Bao, Wayne Wisemandle, Marylynn Baniecki, et al. 2022. “Oral Nirmatrelvir for High-Risk, Nonhospitalized Adults with Covid-19.” The New England Journal of Medicine , 386 (15): 1397–1408

    2022

  6. [14]

    Large-Scale Foundation Model on Single-Cell Transcriptomics

    Hao, Minsheng, Jing Gong, Xin Zeng, Chiming Liu, Yucheng Guo, Xingyi Cheng, Taifeng Wang, Jianzhu Ma, Xuegong Zhang, and Le Song. 2024. “Large-Scale Foundation Model on Single-Cell Transcriptomics.” Nature Methods, 21 (8): 1481–91

    2024

  7. [15]

    Mapping the Genetic Landscape of Human Cells

    Horlbeck, Max A., Albert Xu, Min Wang, Neal K. Bennett, Chong Y . Park, Derek Bogdanoff, Britt Adamson, et al. 2018. “Mapping the Genetic Landscape of Human Cells.” Cell, 174 (4): 953–67.e22

    2018

  8. [16]

    Hsiung, C. C-S, C. M. Wilson, N. A. Sambold, R. Dai, Q. Chen, N. Teyssier, S. Misiukiewicz, et al

  9. [17]

    Additivity Predicts the Efficacy of Most Approved Combination Therapies for Advanced Cancer

    Hwangbo, Haeun, Sarah C. Patterson, Andy Dai, Deborah Plana, and Adam C. Palmer. 2023. “Additivity Predicts the Efficacy of Most Approved Combination Therapies for Advanced Cancer.”Nature Cancer, 4 (12): 1693–1704. 10

    2023

  10. [18]

    Triple–Hormone-Receptor Agonist Retatrutide for Obesity — A Phase 2 Trial

    Jastreboff, Ania M., Lee M. Kaplan, Juan P. Frías, Qiwei Wu, Yu Du, Sirel Gurbuz, Tamer Coskun, Axel Haupt, Zvonko Milicevic, and Mark L. Hartman. 2023. “Triple–Hormone-Receptor Agonist Retatrutide for Obesity — A Phase 2 Trial.” The New England Journal of Medicine , August. h...

    2023

  11. [19]

    Auto-Encoding Variational Bayes

    Kingma, Diederik P., and Max Welling. 2013. “Auto-Encoding Variational Bayes.”http://arxiv.org/ abs/1312.6114

    2013 arXiv

  12. [20]

    Semi-Supervised Classification with Graph Convolutional Networks

    Kipf, Thomas N., and Max Welling. 2016. “Semi-Supervised Classification with Graph Convolutional Networks.” http://arxiv.org/abs/1609.02907

    2016 arXiv

  13. [21]

    DEUP: Direct Epistemic Uncertainty Prediction

    Lahlou, Salem, Moksh Jain, Hadi Nekoei, Victor Ion Butoi, Paul Bertin, Jarrid Rector-Brooks, Maksym Korablyov, and Yoshua Bengio. 2021. “DEUP: Direct Epistemic Uncertainty Prediction.”http://arxiv. org/abs/2102.08501

    2021 arXiv

  14. [22]

    Simple and Scalable Predictive Uncertainty Estimation Using Deep Ensembles

    Lakshminarayanan, Balaji, Alexander Pritzel, and Charles Blundell. 2016. “Simple and Scalable Predictive Uncertainty Estimation Using Deep Ensembles.” http://arxiv.org/abs/1612.01474

    2016 arXiv

  15. [23]

    Learning Causal Representations of Single Cells via Sparse Mechanism Shift Modeling

    Lopez, Romain, Nataša Tagasovska, Stephen Ra, Kyunghyn Cho, Jonathan K. Pritchard, and Aviv Regev. 2022. “Learning Causal Representations of Single Cells via Sparse Mechanism Shift Modeling.” http://arxiv.org/abs/2211.03553

    2022 arXiv

  16. [24]

    Predicting Cellular Responses to Complex Perturbations in High- throughput Screens

    Lotfollahi, Mohammad, Anna Klimovskaia Susmelj, Carlo De Donno, Leon Hetzel, Yuge Ji, Ignacio L. Ibarra, Sanjay R. Srivatsan, et al. 2023. “Predicting Cellular Responses to Complex Perturbations in High- throughput Screens.” Molecular Systems Biology, May. https://doi.org/10.1...

    2023 doi

  17. [25]

    Combination Therapy in Combating Cancer

    Mokhtari, Reza Bayat, Tina S. Homayouni, Narges Baluch, Evgeniya Morgatskaya, Sushil Kumar, Bikul Das, and Herman Yeger. 2017. “Combination Therapy in Combating Cancer.”Oncotarget, 8 (23): 38022

    2017

  18. [26]

    Exploring Genetic Interaction Manifolds Constructed from Rich Single-Cell Phenotypes

    Norman, Thomas M., Max A. Horlbeck, Joseph M. Replogle, Alex Y . Ge, Albert Xu, Marco Jost, Luke A. Gilbert, and Jonathan S. Weissman. 2019. “Exploring Genetic Interaction Manifolds Constructed from Rich Single-Cell Phenotypes.” Science, 365 (6455): 786–93

    2019

  19. [27]

    Rationalizing Combination Therapies

    “Rationalizing Combination Therapies.” 2017. Nature Medicine, 23 (10): 1113–1113

    2017

  20. [28]

    Toward a Foundation Model of Causal Cell and Tissue Biology with a Perturbation Cell and Tissue Atlas

    Rood, Jennifer E., Anna Hupalowska, and Aviv Regev. 2024. “Toward a Foundation Model of Causal Cell and Tissue Biology with a Perturbation Cell and Tissue Atlas.” Cell, 187 (17): 4520–45

    2024

  21. [29]

    Predicting Transcriptional Outcomes of Novel Multigene Perturbations with GEARS

    Roohani, Yusuf, Kexin Huang, and Jure Leskovec. 2023. “Predicting Transcriptional Outcomes of Novel Multigene Perturbations with GEARS.” Nature Biotechnology, 42 (6): 927–35

    2023

  22. [30]

    How May GIP Enhance the Therapeutic Efficacy of GLP-1?

    Samms, Ricardo J., Matthew P. Coghlan, and Kyle W. Sloop. 2020. “How May GIP Enhance the Therapeutic Efficacy of GLP-1?” Trends in Endocrinology and Metabolism: TEM , 31 (6): 410–21

    2020

  23. [31]

    Drug Combination Therapy for Emerging Viral Diseases

    Shyr, Zeenat A., Yu-Shan Cheng, Donald C. Lo, and Wei Zheng. 2021. “Drug Combination Therapy for Emerging Viral Diseases.” Drug Discovery Today, 26 (10): 2367

    2021

  24. [32]

    Mapping the Genetic Interaction Network of PARP Inhibitor Response

    Simpson, Danny, Jia Ling, Yangwode Jing, and Britt Adamson. 2023. “Mapping the Genetic Interaction Network of PARP Inhibitor Response.” bioRxiv, August. https://doi.org/10.1101/2023.08.19. 553986

    2023 doi

  25. [33]

    Induction of Pluripotent Stem Cells from Mouse Embryonic and Adult Fibroblast Cultures by Defined Factors

    Takahashi, Kazutoshi, and Shinya Yamanaka. 2006. “Induction of Pluripotent Stem Cells from Mouse Embryonic and Adult Fibroblast Cultures by Defined Factors.” Cell, 126 (4): 663–76

    2006

  26. [34]

    A Versatile CRISPR-Cas13d Platform for Multiplexed Transcriptomic Regulation and Metabolic Engineering in Primary Human T Cells

    Tieu, Victor, Elena Sotillo, Jeremy R. Bjelajac, Crystal Chen, Meena Malipatlolla, Justin A. Guerrero, Peng Xu, et al. 2024. “A Versatile CRISPR-Cas13d Platform for Multiplexed Transcriptomic Regulation and Metabolic Engineering in Primary Human T Cells.” Cell, 187 (5): 1278–95.e20

    2024

  27. [35]

    Submodularity in Data Subset Selection and Active Learning

    Wei, Kai, Rishabh Iyer, and Jeff Bilmes. 2015. “Submodularity in Data Subset Selection and Active Learning.” In International Conference on Machine Learning , 1954–63. PMLR. 11 A Dataset summary We utilized four bulk combinatorial perturbation datasets in our study: Combinator...

    2015

  28. [1024]

    We also used a linear rate scheduler with 10% of training steps used for warm up, and weight_decay = 0. To identify these optimal hyperparameters, we testing hyperparameters across the following ranges: n_epoch: [50, 100, 200, 500, 1000, 2000] batch_size: [512, 1024, 2048, 409...

  29. [2024]

    Engineered CRISPR-Cas12a for Higher-Order Combinatorial Chromatin Perturbations

    “Engineered CRISPR-Cas12a for Higher-Order Combinatorial Chromatin Perturbations.” Nature Biotechnology, May, 1–15

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.