IP Library › Granted Patent US 12,640,230
Granted Patent B2
US 12,640,230 · App. 18/639,146 · Granted May 26, 2026

Active learning for discovering pairwise interactions via representation learning

Inventors: Aniket Rajiv Didolkar (Montreal, CA); Jason Siyanda Hartford (Montreal, CA); Moksh Mukesh Kumar Jain (Montreal, CA)
Assignee: Recursion Pharmaceuticals, Inc.
G16B25/10G16B40/00
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,640,230
App. No.
18/639,146
Granted
May 26, 2026
Kind
B2
Abstract

The present disclosure relates to systems, non-transitory computer-readable media, and methods that a implement a framework for active learning to discover pairwise interactions via representation learning. Indeed, in one or more implementations, the disclosed systems generate a first individual perturbation embedding from a first representation of a first cell exposed to a first perturbation and a second individual perturbation embedding, from a second representation of a second cell exposed to a second perturbation. For instance, the disclosed systems combine the first individual perturbation embedding and the second individual perturbation embedding to determine a predicted pairwise embedding. Moreover, in some instances, the disclosed systems generate a pairwise embedding from a representation of a cell exposed to both the first and second perturbation. Additionally, from comparing the predicted pairwise embedding with the pairwise embedding, the disclosed systems generate a measure of biological interaction of the first and second perturbation.

Claims (71)

1 . A method comprising:

receiving a first digital image of a first biological cell exposed to a first perturbation, wherein the first perturbation comprises a first gene knockout of a first gene of the first biological cell or a first compound treatment of the first biological cell;

generating, utilizing a neural network machine learning model, a first individual perturbation feature vector, from the first digital image of the first biological cell exposed to the first perturbation;

receiving a second digital image of a second biological cell exposed to a second perturbation, wherein the second perturbation comprises a second gene knockout of a second gene of the second biological cell or a second compound treatment of the second biological cell;

generating, utilizing the neural network machine learning model, a second individual perturbation feature vector, from the second digital image of the second biological cell exposed to the second perturbation;

generating a predicted pairwise feature vector for the first biological cell individually exposed to the first perturbation and the second biological cell individually exposed to the second perturbation by combining the first individual perturbation feature vector and the second individual perturbation feature vector;

receiving a third digital image of a third biological cell exposed to a perturbation pair comprising both the first perturbation and the second perturbation;

generating, utilizing the neural network machine learning model, a double perturbation pairwise feature vector from the third digital image of the third biological cell exposed to the perturbation pair comprising both the first perturbation and the second perturbation;

generating a measure of biological interaction between the first perturbation and the second perturbation by comparing the predicted pairwise feature vector from the first digital image and the second digital image with the double perturbation pairwise feature vector from the third digital image, wherein the measure of biological interaction indicates a difference between biological cell response from exposure to the first perturbation and the second perturbation individually relative to biological cell response from exposure to the first perturbation and the second perturbation in combination;

generating an incomplete biological interaction matrix arranged according to a plurality of perturbation pairs, wherein the incomplete biological interaction matrix comprises the measure of biological interaction for the perturbation pair as an entry in the incomplete biological interaction matrix and further comprises additional measures of biological interactions as additional entries in the incomplete biological interaction matrix that correspond to additional perturbation pairs of the plurality of perturbation pairs;

generating, utilizing an active learning model that takes as inputs entries comprising the measure of biological interaction and the additional measures of biological interaction in the incomplete biological interaction matrix, predicted pairwise perturbation interaction scores corresponding to incomplete entries in the incomplete biological interaction matrix;

based on the predicted pairwise perturbation interaction scores, selecting additional perturbations for experimentation from the incomplete biological interaction matrix, wherein the additional perturbations for experimentation correspond to incomplete entries in the incomplete biological interaction matrix; and

performing a downstream experiment for a fourth biological cell by perturbing the fourth biological cell according to the additional perturbations selected from the incomplete biological interaction matrix.

2 . The method of claim 1 , wherein:

receiving the first digital image comprises receiving a first captured image of the first biological cell with a first modified cell phenotype resulting from the first gene knockout of the first gene; and

receiving the second digital image comprises receiving a second captured image of the second biological cell with a second modified cell phenotype resulting from the second gene knockout of the second gene.

3 . The method of claim 1 , wherein determining the predicted pairwise feature vector for the first perturbation and the second perturbation comprises:

determining a first normalized feature vector from the first individual perturbation feature vector and a control feature vector and a second normalized feature vector from the second individual perturbation feature vector and the control feature vector; and

combining the first normalized feature vector and the second normalized feature vector to determine the predicted pairwise feature vector for the first perturbation and the second perturbation.

4 . The method of claim 1 , further comprising generating the double perturbation pairwise feature vector from the third digital image portraying the third biological cell exposed to a double gene knockout of the first gene and the second gene in the third biological cell.

5 . The method of claim 1 , wherein generating the incomplete biological interaction matrix comprises generating a matrix that includes pairwise perturbation experiment data corresponding to actual pairwise perturbations and predictions of measures of biological interactions corresponding to perturbation pairs.

6 . The method of claim 5 , further comprises generating, utilizing the active learning model that comprises a pairwise prediction model, the predicted pairwise perturbation interaction scores for individual perturbations and corresponding information gain predictions based on the incomplete biological interaction matrix, wherein the corresponding information gain predictions indicate an amount of potential increase in knowledge relative to the incomplete biological interaction matrix in response to a selection of the additional perturbations for experimentation.

7 . The method of claim 6 , wherein utilizing the active learning model comprises an active-matrix completion algorithm to select a first entry from a plurality of entries of the incomplete biological interaction matrix based on the predicted pairwise perturbation interaction scores for the individual perturbations and the corresponding information gain predictions.

8 . The method of claim 1 , further comprising:

based on performing the downstream experiment for the fourth biological cell, determining additional measure of biological interaction for the additional perturbations selected from the incomplete biological interaction matrix; and

updating the incomplete biological interaction matrix to include the additional measure of biological interaction for the additional perturbations.

9 . A method comprising:

receiving a first digital image of a first biological cell exposed to a first gene knockout of a first gene of the first biological cell;

generating, utilizing a neural network machine learning model, a first individual perturbation feature vector, from the first digital image of the first biological cell exposed to the first gene knockout;

receiving a second digital image of a second biological cell exposed to a second gene knockout of a second gene of the second biological cell;

generating, utilizing the neural network machine learning model, a second individual perturbation feature vector, from the second digital image of the second biological cell exposed to the second gene knockout;

generating a predicted pairwise feature vector for the first biological cell individually exposed to the first gene knockout and the second biological cell individually exposed to the second gene knockout by combining the first individual perturbation feature vector and the second individual perturbation feature vector;

receiving a third digital image of a third biological cell exposed to a gene knockout pair comprising both the first gene knockout and the second gene knockout;

generating, utilizing the neural network machine learning model, a double perturbation pairwise feature vector from the third digital image of the third biological cell exposed to the gene knockout pair comprising both the first gene knockout and the second gene knockout;

generating a measure of biological interaction between the first gene knockout and the second gene knockout by comparing the predicted pairwise feature vector from the first digital image and the second digital image with the double perturbation pairwise feature vector from the third digital image, wherein the measure of biological interaction indicates a difference between biological cell response from exposure to the first gene knockout and the second gene knockout individually relative to biological cell response from exposure to the first gene knockout and the second gene knockout in combination;

generating an incomplete biological interaction matrix arranged according to a plurality of gene knockout pairs, wherein the incomplete biological interaction matrix comprises the measure of biological interaction for the gene knockout pair as an entry in the incomplete biological interaction matrix and further comprises additional measures of biological interactions as additional entries in the incomplete biological interaction matrix that correspond to additional gene knockout pairs of the plurality of gene knockout pairs;

generating, utilizing an active learning model that takes as inputs entries comprising the measure of biological interaction and the additional measures of biological interaction in the incomplete biological interaction matrix, predicted pairwise perturbation interaction scores corresponding to incomplete entries in the incomplete biological interaction matrix;

based on the predicted pairwise perturbation interaction scores, selecting additional gene knockouts for experimentation from the incomplete biological interaction matrix, wherein the additional gene knockouts for experimentation correspond to incomplete entries in the incomplete biological interaction matrix; and

performing a downstream experiment for a fourth biological cell by applying the additional gene knockouts selected from the incomplete biological interaction matrix to the fourth biological cell.

10 . The method of claim 9 , wherein:

receiving the first digital image comprises receiving a first captured image of the first biological cell with a first modified cell phenotype resulting from the first gene knockout of the first gene; and

receiving the second digital image comprises receiving a second captured image of the second biological cell with a second modified cell phenotype resulting from the second gene knockout of the second gene.

11 . The method of claim 9 , wherein determining the predicted pairwise feature vector for the first gene knockout and the second gene knockout comprises:

determining a first normalized feature vector from the first individual perturbation feature vector and a control feature vector and a second normalized feature vector from the second individual perturbation feature vector and the control feature vector; and

combining the first normalized feature vector and the second normalized feature vector to determine the predicted pairwise feature vector for the first gene knockout and the second gene knockout.

12 . The method of claim 9 , further comprising generating the double perturbation pairwise feature vector from the third digital image portraying the third biological cell exposed to a double gene knockout of the first gene and the second gene in the third biological cell.

13 . The method of claim 9 , wherein generating the incomplete biological interaction matrix comprises generating a matrix that includes pairwise perturbation experiment data corresponding to actual pairwise perturbations and predictions of measures of biological interactions corresponding to perturbation pairs.

14 . The method of claim 9 , further comprising generating, utilizing the active learning model that comprises a pairwise prediction model, the predicted pairwise perturbation interaction scores for individual gene knockouts and corresponding information gain predictions based on the incomplete biological interaction matrix, wherein the corresponding information gain predictions indicate an amount of potential increase in knowledge relative to the incomplete biological interaction matrix in response to a selection of the additional gene knockouts for experimentation.

15 . The method of claim 14 , wherein utilizing the active learning model comprises utilizing an active-matrix completion algorithm to select a first entry from a plurality of entry pairs of the incomplete biological interaction matrix based on the predicted pairwise perturbation interaction scores for the individual gene knockouts and the corresponding information gain predictions.

16 . A method comprising:

capturing, utilizing one or more digital cameras, a first digital image of a first biological cell exposed to a first perturbation, wherein the first perturbation comprises a first gene knockout of a first gene of the first biological cell or a first compound treatment of the first biological cell;

generating, utilizing a neural network machine learning model, a first individual perturbation feature vector, from the first digital image of the first biological cell exposed to the first perturbation;

capturing, utilizing the one or more digital cameras, a second digital image of a second biological cell exposed to a second perturbation, wherein the second perturbation comprises a second gene knockout of a second gene of the second biological cell or a second compound treatment of the second biological cell;

generating, utilizing the neural network machine learning model, a second individual perturbation feature vector, from the second digital image of the second biological cell exposed to the second perturbation;

generating a predicted pairwise feature vector for the first biological cell individually exposed to the first perturbation and the second biological cell individually exposed to the second perturbation by combining within a machine learning feature space, the first individual perturbation feature vector and the second individual perturbation feature vector;

capturing, utilizing the one or more digital cameras, a third digital image of a third biological cell exposed to a perturbation pair comprising both the first perturbation and the second perturbation;

generating, utilizing the neural network machine learning model, a double perturbation pairwise feature vector from the third digital image of the third biological cell exposed to the perturbation pair comprising both the first perturbation and the second perturbation;

generating a measure of biological interaction between the first perturbation and the second perturbation by comparing the predicted pairwise feature vector from the first digital image and the second digital image with the double perturbation pairwise feature vector from the third digital image, wherein the measure of biological interaction indicates a difference between biological cell response from exposure to the first perturbation and the second perturbation individually relative to biological cell response from exposure to the first perturbation and the second perturbation in combination;

generating an incomplete biological interaction matrix arranged according to a plurality of perturbation pairs, wherein the incomplete biological interaction matrix comprises the measure of biological interaction for the perturbation pair as an entry in the incomplete biological interaction matrix and further comprises additional measures of biological interactions as additional entries in the incomplete biological interaction matrix that correspond to additional perturbation pairs of the plurality of perturbation pairs;

generating, utilizing an active learning model that takes as inputs entries comprising the measure of biological interaction and the additional measures of biological interaction in the incomplete biological interaction matrix, predicted pairwise perturbation interaction scores corresponding to incomplete entries in the incomplete biological interaction matrix;

based on the predicted pairwise perturbation interaction scores, selecting additional perturbations for experimentation from the incomplete biological interaction matrix, wherein the additional perturbations for experimentation correspond to incomplete entries in the incomplete biological interaction matrix;

performing a downstream experiment for a fourth biological cell by perturbing, utilizing an experimental device, the fourth biological cell according to the additional perturbations selected from the incomplete biological interaction matrix; and

capturing, utilizing the one or more digital cameras, a fourth digital image of the fourth biological cell exposed to the additional perturbations selected from the incomplete biological interaction matrix.

17 . The method of claim 16 , wherein:

receiving the first digital image comprises receiving a first captured image of the first biological cell with a first modified cell phenotype resulting from the first gene knockout of the first gene; and

receiving the second digital image comprises receiving a second captured image of the second biological cell with a second modified cell phenotype resulting from the second gene knockout of the second gene.

18 . The method of claim 16 , wherein determining the predicted pairwise feature vector for the first perturbation and the second perturbation comprises:

determining a first normalized feature vector from the first individual perturbation feature vector and a control feature vector and a second normalized feature vector from the second individual perturbation feature vector and the control feature vector; and

combining the first normalized feature vector and the second normalized feature vector to determine the predicted pairwise feature vector for the first perturbation and the second perturbation.

19 . The method of claim 16 , further comprising generating the double perturbation pairwise feature vector from the third digital image portraying the third biological cell exposed to a double gene knockout of the first gene and the second gene in the third biological cell.

20 . The method of claim 16 , wherein generating the incomplete biological interaction matrix comprises generating a matrix that includes pairwise perturbation experiment data corresponding to actual pairwise perturbations and predictions of measures of biological interactions corresponding to perturbation pairs.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 20, 2024
From: DIDOLKAR, ANIKET RAJIV; HARTFORD, JASON SIYANDA; JAIN, MOKSH MUKESH KUMAR
To: RECURSION PHARMACEUTICALS, INC.
Reel/Frame 067786/0974 →
Continuity (2)
Provisional Application 63551314 · Feb 8, 2024
Related Publication 20250259705A1 · Aug 14, 2025
References Cited (71)
US 10769501B1 · Ando et al. · 2020 [cited by applicant]
US 12073638B1 · Lazar et al. · 2024 [cited by examiner]
US 20190085324A1 · Regev et al. · 2019 [cited by examiner]
US 20200362334A1 · Regev et al. · 2020 [cited by examiner]
US 20210071256A1 · Quigley et al. · 2021 [cited by examiner]
US 20230170040A1 · Hauck · 2023 [cited by applicant]
WO 2022087540A1 · 2022 [cited by applicant]
Yu, Hengshi. “Deep Generative Models for Single-Cell Perturbation Experiments.” ProQuest Dissertations & Theses, 2022. Print. (Year: 2022). [cited by examiner]
Peidli, Stefan et al. “scPerturb: Harmonized Single-Cell Perturbation Data.” Nature methods 21.3 (2024): 531-540. Web. (Year: 2024). [cited by examiner]
Chandrasekaran, Srinivas Niranj et al. “Image-Based Profiling for Drug Discovery: Due for a Machine-Learning Upgrade?” Nature reviews. Drug discovery 20.2 (2021): 145-159. Web. (Year: 2021). [cited by examiner]
Yu, Hengshi. PerturbNet Predicts Single-Cell Responses to Unseen Chemical and Genetic Perturbations. NewsRX LLC, 2022. Print. (Year: 2022). [cited by examiner]
Bunne, Charlotte et al. “Learning Single-Cell Perturbation Responses Using Neural Optimal Transport.” Nature methods 20.11 (2023): 1759-1768. Web. (Year: 2023). [cited by examiner]
Alessandro Palma, Fabian J. Theis, Mohammad Lotfollahi bioRxiv 2023.07.17.549216; doi: https://doi.org/10.1101/2023.07.17.549216 (Year: 2023). [cited by examiner]
Sun H, Murphy RF. Evaluation of categorical matrix completion algorithms: toward improved active learning for drug discovery. Bioinformatics. Oct. 25, 2021;37(20):3538-3545 (Year: 2021). [cited by examiner]
Osorio, Daniel et al. “scTenifoldKnk: An Efficient Virtual Knockout Tool for Gene Function Predictions via Single-Cell Gene Regulatory Network Perturbation.” Patterns (New York, N.Y.) 3.3 (2022): 100434-100434. Web. (Ye… [cited by examiner]
Caldera, Michael et al. “Mapping the Perturbome Network of Cellular Perturbations.” Nature communications 10.1 (2019): 5140-14. Web. (Year: 2019). [cited by examiner]
Roohani, Yusuf, Kexin Huang, and Jure Leskovec. “Predicting Transcriptional Outcomes of Novel Multigene Perturbations with GEARS.” Nature biotechnology 42.6 (2023): 927-935. Web. (Year: 2023). [cited by examiner]
Feldman, David et al. “Pooled Genetic Perturbation Screens with Image-Based Phenotypes.” Nature protocols 17.2 (2022): 476-512. Web. (Year: 2022). [cited by examiner]
Cheng, Junyun et al. “Massively Parallel CRISPR-Based Genetic Perturbation Screening at Single-Cell Resolution.” Advanced science 10.4 (2023): e2204484-n/a. Web. (Year: 2023). [cited by examiner]
A. Foster. Variational, Monte Carlo and Policy-Based Approaches to Bayesian Experimental Design, PhD thesis, University of Oxford, 2021. [cited by applicant]
A. Karbasi, V. Mirrokni, and M. Shadravan. Parallelizing Thompson Sampling. Advances in Neural Information Processing Systems, 34:10535-10548, 2021. [cited by applicant]
A. Mehrjou, A. Soleymani, A. Jesson, P. Notin, Y. Gal, S. Bauer, and P. Schwab. GeneDisco: A Benchmark for Experimental Design in Drug Discovery. In International Conference on Learning Representations, arXiv:2110.11875… [cited by applicant]
A. Pacchiano, D. Wulsin, R. A. Barton, and L. Voloch. Neural Design for Genetic Perturbation Experiments. In The Eleventh International Conference on Learning Representations, 2023b. [cited by applicant]
A. Pacchiano, J. Lee, and E. Brunskill. Experiment Planning with Function Approximation. In Thirty-seventh Conference on Neural Information Processing Systems, 2023a. [cited by applicant]
A. Zanette, K. Dong, J. N. Lee, and E. Brunskill. Design of Experiments for Stochastic Contextual Linear Bandits. Advances in Neural Information Processing Systems, 34:22720-22731, 2021. [cited by applicant]
C. Lyle, A. Mehrjou, P. Notin, A. Jesson, S. Bauer, Y. Gal, and p. Schwab. DiscoBAX: Discovery of Optimal Intervention Sets in Genomic Experiment Design. In International Conference on Machine Learning, pp. 23170-23189.… [cited by applicant]
D. Kuzuoglu-Ozturk, Z. Hu, M. Rama, E. Devericks, J. Weiss, G. G. Chiang, S. T. Worland, S. E. Brenner, H. Goodarzi, L. A. Gilbert, et al. Revealing molecular pathways for cancer cell fitness through a genetic screen of… [cited by applicant]
D. Phan, N. Pradhan, and M. Jankowiak. Composable Effects for Flexible and Accelerated Probabilistic Programming in NumPyro. arXiv preprint arXiv:1912.11554v1, Dec. 24, 2019. [cited by applicant]
D. Russo and B. Van Roy. An Information-Theoretic Analysis of Thompson Sampling. The Journal of Machine Learning Research, 17(1):2442-2471, 2016. [cited by applicant]
D. Szklarczyk, A. L. Gable, K. C. Nastou, D. Lyon, R. Kirsch, S. Pyysalo, N. T. Doncheva, M. Legeay, T. Fang, P. Bork, et al. The string database in 2021: customizable protein-protein networks, and functional characteri… [cited by applicant]
D. V. Lindley. On a Measure of the Information Provided by an Experiment. The Annals of Mathematical Statistics, 27(4):986-1005, 1956. [cited by applicant]
D. Wingate and T. Weber. Automated Variational Inference in Probabilistic Programming. arXiv preprint arXiv:1301.1299v1, Jan. 7, 2013. [cited by applicant]
D. Xun, R.Wang, X. Zhang, and Y.Wang. Microsnoop: A Generalized Tool for Unbiased Representation of Diverse Microscopy Images. bioRxiv, 2023. doi: 10.1101/2023.02.25.530004. URL https://www.biorxiv.org/content/early/202… [cited by applicant]
E. Bingham, J. P. Chen, M. Jankowiak, F. Obermeyer, N. Pradhan, T. Karaletsos, R. Singh, P. A. Szerlip, P. Horsfall, and N. D. Goodman. Pyro: Deep Universal Probabilistic Programming. J. Mach. Learn. Res., 20:28:1-28:6,… [cited by applicant]
E. G. Ryan, C. C. Drovandi, J. M. McGree, and A. N. Pettitt. A Review of Modern Computational Algorithms for Bayesian Optimal Design. International Statistical Review, 84(1):128-154, 2016. [cited by applicant]
G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger. Densely Connected Convolutional Networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 4700-4708, 2017. [cited by applicant]
G. Roeder, L. Metz, and D. Kingma. On Linear Identifiability of Learned Representations. In M. Meila and T. Zhang, editors, Proceedings of the 38th International Conference on Machine Learning, vol. 139 of Proceedings o… [cited by applicant]
H. Robbins. Some Aspects of the Sequential Design of Experiments. 1952. [cited by applicant]
J. A. Doudna and E. Charpentier. The new frontier of genome engineering with CRISPR-Cas9. Science, 346(6213):1258096, 2014. [cited by applicant]
K. Drew, J. B. Wallingford, and E. M. Marcotte. hu. MAP 2.0: integration of over 15,000 proteomic experiments builds a global compendium of human multiprotein assemblies. Molecular systems biology, 17(5):e10016, 2021. [cited by applicant]
K. Huang, R. Lopez, J.-C. Hutter, T. Kudo, A. Rios, and A. Regev. Sequential Optimal Experimental Design of Perturbation Screens Guided by Multi-modal Priors. bioRxiv, pp. 2023-12, 2023. [cited by applicant]
L. Licata, P. Lo Surdo, M. Iannuccelli, A. Palma, E. Micarelli, L. Perfetto, D. Peluso, A. Calderone, L. Castagnoli, and G. Cesareni. SIGNOR 2.0, the SIGnaling Network Open Resource 2.0: 2019 update. Nucleic Acids Resea… [cited by applicant]
M. Bereket and T. Karaletsos. Modelling Cellular Perturbations with the Sparse Additive Mechanism Shift Variational Autoencoder. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL https://o… [cited by applicant]
M. Giurgiu, J. Reinhard, B. Brauner, I. Dunger-Kaltenbach, G. Fobo, G. Frishman, C. Montrone, and A. Ruepp. CORUM: the comprehensive resource of mammalian protein complexes—2019. Nucleic Acids Research, 47(D1):D559-D563… [cited by applicant]
M. Jinek, K. Chylinski, I. Fonfara, M. Hauer, J. A. Doudna, and E. Charpentier. A programmable dual RNA-guided DNA endonuclease in adaptive bacterial immunity. Science, 337(6096):816-821, 2012. [cited by applicant]
M. Lotfollahi, A. Klimovskaia Susmelj, C. De Donno, L. Hetzel, Y. Ji, I. L. Ibarra, S. R. Srivatsan, M. Naghipourfar, R. M. Daza, B. Martin, J. Shendure, J. L. McFaline-Figueroa, P. Boyeau, F. A. Wolf, N. Yakubova, S. G… [cited by applicant]
M. M. Fay, O. Kraus, M. Victors, L. Arumugam, K. Vuggumudi, J. Urbanik, K. Hansen, S. Celik, N. Cernek, G. Jagannathan, et al. RxRx3: Phenomics Map of Biology. bioRxiv, pp. 2023-02, 2023. [cited by applicant]
M. Sypetkowski, M. Rezanejad, S. Saberian, O. Kraus, J. Urbanik, J. Taylor, B. Mabey, M. Victors, J. Yosinski, A. R. Sereshkeh, et al. RxRx1: A Dataset for Evaluating Experimental Batch Correction Methods. In Proceeding… [cited by applicant]
N. Moshkov, M. Bornholdt, S. Benoit, M. Smith, C. McQuin, A. Goodman, R. A. Senft, Y. Han, M. Babadi, P. Horvath, et al. Learning representations for image-based profiling of perturbations. Biorxiv, pp. 2022-08, 2022. [cited by applicant]
N. Srinivas, A. Krause, S. Kakade, and M. Seeger. Gaussian Process Optimization in the Bandit Setting: No. Regret and Experimental Design. In Proceedings of the 27th International Conference on International Conference … [cited by applicant]
O. Kraus, K. Kenyon-Dean, S. Saberian, M. Fallah, P. McLean, J. Leung, V. Sharma, A. Khan, J. Balakrishnan, S. Celik, et al. Masked Autoencoders are Scalable Learners of Cellular Corphology. arXiv preprint arXiv:2309.16… [cited by applicant]
P. Sebastiani and H. P. Wynn. Maximum entropy sampling and optimal Bayesian experimental design. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 62(1):145-157, 2000. [cited by applicant]
R. Combes, M. S. Talebi Mazraeh Shahi, A. Proutiere, et al. Combinatorial Bandits Revisited. Advances in neural information processing systems, 28, 2015. [cited by applicant]
R. Garnett, Y. Krishnamurthy, X. Xiong, J. Schneider, and R. Mann. Bayesian Optimal Active Search and Surveying. In Proceedings of the 29th International Conference on International Conference on Machine Learning, pp. 8… [cited by applicant]
R. Lopez, N. Tagasovska, S. Ra, K. Cho, J. Pritchard, and A. Regev. Learning Causal Representations of Single Cells via Sparse Mechanism Shift Modeling. In M. van der Schaar, C. Zhang, and D. Janzing, editors, Proceedin… [cited by applicant]
R. Ranganath, S. Gerrish, and D. Blei. Black Box Variational Inference. In Artificial intelligence and statistics, pp. 814-822. PMLR, 2014. [cited by applicant]
S. M. Nijman. Synthetic lethality: General principles, utility and detection using genetic screens in human cells. FEBS Letters, 585(1):1-6, 2011. ISSN 0014-5793. [cited by applicant]
S. N. Chandrasekaran, J. Ackerman, E. Alix, D. M. Ando, J. Arevalo, M. Bennion, N. Boisseau, A. Borowa, J. D. Boyd, L. Brino, P. J. Byrne, et al. JUMP Cell Painting dataset: morphological impact of 136,000 chemical and … [cited by applicant]
T. Desautels, A. Krause, and J. W. Burdick. Parallelizing Exploration-Exploitation Tradeoffs in Gaussian Process Bandit Optimization. Journal of Machine Learning Research, 15:3873-3923, 2014. [cited by applicant]
T. Lattimore and C. Szepesvári. Bandit Algorithms. Cambridge University Press, 2020. [cited by applicant]
T. Rainforth, A. Foster, D. R. Ivanova, and F. B. Smith. Modern Bayesian Experimental Design. arXiv preprint arXiv:2302.14545v2, Nov. 29, 2023. [cited by applicant]
T. Rainforth, R. Cornish, H. Yang, A. Warrington, and F. Wood. On Nesting Monte Carlo Estimators. In International Conference on Machine Learning, pp. 4267-4276. PMLR, 2018. [cited by applicant]
V. Kanade, H. B. McMahan, and B. Bryan. Sleeping Experts and Bandits with Stochastic Action Availability and Adversarial Rewards. In Artificial Intelligence and Statistics, pp. 272-279. PMLR, 2009. [cited by applicant]
W. Chen, Y. Wang, and Y. Yuan. Combinatorial Multi-Armed Bandit: General Famework, Results and Applications. In International conference on machine learning, pp. 151-159. PMLR, 2013. [cited by applicant]
W. R. Thompson. On the Likelihood that One Unknown Probability Exceeds Another in View of the Evidence of Two Samples. Biometrika, 25(3-4):285-294, 1933. [cited by applicant]
Z. Dai, Q. P. Nguyen, S. S. Tay, D. Urano, R. Leong, B. K. H. Low, and P. Jaillet. Batch Bayesian Optimization for Replicable Experimental Design. arXiv preprint arXiv:2311.01195v1, Nov. 2, 2023. [cited by applicant]
Z. Wang, L. Gui, J. Negrea, and V. Veitch. Concept Algebra for (Score-Based) Text-Controlled Generative Models. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL https://openreview.net/for… [cited by applicant]
Z. Xu, E. Shim, A. Tewari, and P. Zimmerman. Adaptive Sampling for Discovery. Advances in Neural Information Processing Systems, 35:1114-1126, 2022. [cited by applicant]
GitHub - Google/JAX Composable transformation of Phython+NumPy programs. 6 pages [retrieved on Jun. 7, 2024]. Retrieved from the Internet: https://github.com/jax-ml/jax. [cited by applicant]
Mark-Anthony Bray et al. Cell Painting, a high-content image-based assay for morphological profiling using multiplexed fluorescent dyes. Nature Protocols, 11(9):1757-1774, 2016. [cited by applicant]
International Search Report and Written Opinion as received in PCT/US2025/011845 dated Apr. 21, 2025. [cited by applicant]