Causal Cellular Context Transfer Learning (C3TL): An Efficient Architecture for Prediction of Unseen Perturbation Effects

Michael Scholkemper; Sach Mukherjee

cs.LGMar 13, 2026

Causal Cellular Context Transfer Learning (C3TL): An Efficient Architecture for Prediction of Unseen Perturbation Effects

Authors: Michael Scholkemper, Sach Mukherjee

Organizations: Statistics and Machine Learning, German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany. · Statistics and Machine Learning, German Center for Neurodegenerative Diseases (DZNE) Bonn, Germany, University of Bonn, Germany & University of Cambridge, United Kingdom.

Abstract

Predicting the effects of chemical and genetic perturbations on quantitative cell states is a central challenge in computational biology, molecular medicine and drug discovery. Recent work has leveraged large-scale single-cell data and massive foundation models to address this task. However, such computational resources and extensive datasets are not always accessible in academic or clinical settings, hence limiting utility. Here we propose a lightweight framework for perturbation effect prediction that exploits the structured nature of biological interventions and specific inductive biases/invariances. Our approach leverages available information concerning perturbation effects to allow generalization to novel contexts and requires only widely-available bulk molecular data. Extensive testing, comparing predictions of context-specific perturbation effects against real, large-scale interventional experiments, demonstrates accurate prediction in new contexts. The proposed approach is competitive with SOTA foundation models but requires simpler data, much smaller model sizes and less time. Focusing on robust bulk signals and efficient architectures, we show that accurate prediction of perturbation effects is possible without proprietary hardware or very large models, hence opening up ways to leverage causal learning approaches in biomedicine generally.

Figures & tables

Figure 1: Proposed architecture. The model compresses the high-dimensional gene expression input

\{x^{\gamma}_{p}\}_{\gamma}

from multiple contexts (indexed by

\gamma

) but under the same perturbation

p

into a lower-dimensional latent representation

\hat{z}_{p}

of the perturbation itself. Similarly, it forms a representation

\hat{\psi}_{c}

of the context from all the different perturbations (indexed by

\pi

) observed in the specific context

c

. From these latent representations, it then reconstructs the original gene expression

\hat{x}^{c}_{p}

using a decoder

\Gamma

	Replogle			Parse			Tahoe
Method	Correlation ( $\uparrow$ )	MSE ( $\downarrow$ )	Epoch	Correlation ( $\uparrow$ )	MSE ( $\downarrow$ )	Epoch	Correlation ( $\uparrow$ )	MSE ( $\downarrow$ )	Epoch
C3TL (Ours)	0.491 $\pm$ 0.031	0.016 $\pm$ 0.005	0.25s	0.670 $\pm$ 0.202	0.019 $\pm$ 0.018	0.11s	0.777 $\pm$ 0.074	0.002 $\pm$ 0.000	1.7s
State	0.474 $\pm$ 0.047	0.018 $\pm$ 0.006	10s	0.608 $\pm$ 0.099	0.104 $\pm$ 0.028	5.4s	0.778 $\pm$ 0.069	0.004 $\pm$ 0.003	59s
CPA	0.008 $\pm$ 0.026	0.028 $\pm$ 0.011	8.9s	0.000 $\pm$ 0.000	0.589 $\pm$ 0.102	3.8s	0.13 $\pm$ 0.015	0.010 $\pm$ 0.003	87s
Mean Baseline	0.370 $\pm$ 0.041	62.453 $\pm$ 24.680	-	0.627 $\pm$ 0.151	4.602 $\pm$ 1.400	-	0.725 $\pm$ 0.083	1.684 $\pm$ 0.345	-
Closest Cell Baseline	0.319 $\pm$ 0.059	90.141 $\pm$ 30.987	-	0.523 $\pm$ 0.147	0.926 $\pm$ 0.641	-	0.630 $\pm$ 0.067	0.682 $\pm$ 0.300	-

Table 1: Performance of the models on perturbation datasets. The table shows the mean of the Pearson Correlation of the prediction of the individual algorithms with the true target for each gene as well as the Mean Squared Error. Metrics are reported as the mean (

\pm

standard deviation) over a 5-fold cross validation (4-fold for the Replogle dataset). Best and second best performing methods are colored in blue and gray respectively. In all experiments, metrics are only computed on interventions that have not been seen by the models in the target context.

Figure 2: Scatter plot of model predictions. Comparison of model outputs versus true target values for the Tahoe-100 dataset. The dashed line represents perfect prediction.

Figure 3: Sensitivity to data scarcity. Performance evaluation under varying training data availability. Panels (left to right) display regimes with 43, 20, 10, and 5 randomly sampled training contexts. The x-axis represents the number of interventions in the test context available for adaptation (implying the complement was used for testing). This ticks correspond to (from left to right)

80\%

50\%

30\%

20\%

10\%

5\%

1\%

of the data available in the test contexts in total. The y-axis displays the Pearson Correlation between prediction and target (

\pm

standard deviation across 5 test contexts). In all experiments, metrics are only computed on interventions that have not been seen by the models in the target context.

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

	Tahoe-100	Replogle	Parse
Type	Compound	Genetic	Signaling
# Contexts	48	4	24
# Pert. / Context	1138	1677	90
Reference	[ 20 ]	[ 2 , 3 ]	[ 21 ]

Appendix

Table 2: Overview of datasets used for evaluation. The datasets cover different perturbation modalities including compound, genetic, and signaling perturbations.

Transfer Learning

Transfer learning studies how models reuse knowledge across tasks, domains, or modalities when target data or computation are limited.

See all 417 papers

MapPFN: Learning Causal Perturbation Maps in Context
Jan 28, 2026Marvin Sextro, Weronika Kłos, Gabriel DernbachSingle-Cell Perturbation Modeling Causal Intervention
Machine Learning Group, Technische Universität Berlin, Berlin, Germany · Berlin Institute for the Foundations of Learning and Data (BIFOLD) · Aignostics, Berlin, Germany +1
Learning Latent Dynamical Causal Processes for Single-Cell Perturbation Prediction
May 25, 2026Wenkang Jiang, Yuhang Liu, Erdun Gao +3Causal Representation Learning Latent Dynamics Modeling
AIML, Adelaide University Adelaide, Australia · Responsible AI Research Centre AIML, Adelaide University Adelaide, Australia · Monash University Melbourne, Australia +1
Differentiable Retrieval-Augmented Generation for Predicting Cellular Responses to Gene Perturbation
Mar 7, 2026Andrea Giuseppe Di Francesco, Andrea Rubbi, Rishabh Jain +1Single-Cell Perturbation Modeling Retrieval-Augmented Generation
University of Cambridge Wellcome Sanger Institute

Jan 28, 2026J/K move · Enter open · S save

MapPFN: Learning Causal Perturbation Maps in Context

Marvin Sextro, Weronika Kłos, Gabriel Dernbach

Machine Learning Group, Technische Universität Berlin, Berlin, Germany · Berlin Institute for the Foundations of Learning and Data (BIFOLD) · Aignostics, Berlin, Germany+1

Single-Cell Perturbation Modeling Causal Intervention Prior-Data Fitted Networks

Open paper AskPDF arXiv

Planning effective interventions in biological systems requires treatment-effect models that adapt to unseen biological contexts by identifying their specific underlying mechanisms. Yet single-cell perturbation datasets span only a handful of biological contexts, and existing methods cannot leverage new interventional evidence at inference time to adapt beyond their training data. To meta-learn a perturbation effect estimator, we present MapPFN, a prior-data fitted network (PFN) pre-trained on a synthetic biological prior with causal interventions, decoupling pre-training from limited wet-lab data. Unlike existing methods, MapPFN uses in-context learning to map a sequence of experiments to a post-perturbation distribution, enabling a single pre-trained model to adapt to new datasets and gene sets. Our results show that combining synthetic pre-training with fine-tuning on real single-cell data improves perturbation predictions across biological contexts. The code, model and data are available at https://marvinsxtr.github.io/MapPFN.

Abstract

Transfer Learning

Explore similar work

MapPFN: Learning Causal Perturbation Maps in Context

Learning Latent Dynamical Causal Processes for Single-Cell Perturbation Prediction

Differentiable Retrieval-Augmented Generation for Predicting Cellular Responses to Gene Perturbation