publications
Publications and preprints in reverse chronological order.
2026
- EPTCSComposable Uncertainty in Symmetric Monoidal Categories for Design ProblemsMarius Furter, Yujun Huang, and Gioele ZardiniElectronic Proceedings in Theoretical Computer Science, Mar 2026
Applied category theory often studies symmetric monoidal categories (SMCs) whose morphisms represent open systems. These structures naturally accommodate complex wiring patterns, leveraging (co)monoidal structures for splitting and merging wires, or compact closed structures for feedback. A key example is the compact closed SMC of design problems (DP), which enables a compositional approach to co-design in engineering. However, in practice, the systems of interest may not be fully known. Recently, Markov categories have emerged as a powerful framework for modeling uncertain processes. In this work, we demonstrate how to integrate this perspective into the study of open systems while preserving consistency with the underlying SMC structure. To this end, we employ the change-of-base construction for enriched categories, replacing the morphisms of a symmetric monoidal V-category C with parametric maps A -> C(X,Y) in a Markov category induced by a symmetric monoidal monad. This results in a symmetric monoidal 2-category N*C with the same objects as C and reparametrization 2-cells. By choosing different monads, we capture various types of uncertainty. The category underlying C embeds into N*C via a strict symmetric monoidal functor, allowing (co)monoidal and compact closed structures to be transferred. Applied to DP, this construction leads to categories of practical relevance, such as parametrized design problems for optimization, and parametrized distributions of design problems for decision theory and Bayesian learning.
2025
- PreprintOn Composable and Parametric Uncertainty in Systems Co-DesignYujun Huang, Marius Furter, and Gioele ZardiniApr 2025
Optimizing the design of complex systems requires navigating interdependent decisions, heterogeneous components, and multiple objectives. Our monotone theory of co-design offers a compositional framework for addressing this challenge, modeling systems as Design Problems (DPs), representing trade-offs between functionalities and resources within partially ordered sets. While current approaches model uncertainty using intervals, capturing worst- and best-case bounds, they fail to express probabilistic notions such as risk and confidence. These limitations hinder the applicability of co-design in domains where uncertainty plays a critical role. In this paper, we introduce a unified framework for composable uncertainty in co-design, capturing intervals, distributions, and parametrized models. This extension enables reasoning about risk-performance trade-offs and supports advanced queries such as experiment design, learning, and multi-stage decision making. We demonstrate the expressiveness and utility of the framework via a numerical case study on the uncertainty-aware co-design of task-driven Unmanned Aerial Vehicle (UAV).
- JNOMultivariable Prediction Model for Suspected Ocular Myasthenia Gravis: Development and ValidationArmin Handzic, Marius P. Furter, Brigitte C. Messmer, and 5 more authorsJournal of Neuro-Ophthalmology, Apr 2025
Objective: Diagnosing ocular myasthenia gravis (OMG) remains challenging despite recent diagnostic advances. We addressed this challenge by developing and validating a multivariable prediction model that estimates the OMG probability given the results of any partial selection of available diagnostic tests. Methods: The source data for our model were retrieved from our blinded prospective diagnostic accuracy study at the University Hospital Zurich (USZ). Patients with ptosis and/or diplopia whose presentation was suspicious for OMG underwent comprehensive diagnostic testing. An independent neuromuscular specialist made the final diagnosis. These data were used to fit and validate a Bayesian network model against additional retrospective USZ and the University of Toronto (UoT) patient data. The primary outcome was to predict the likelihood of a positive OMG diagnosis given the available diagnostic tests. For any set of tests, the model returns an OMG probability together with 95% credible intervals, indicating the prediction uncertainty. Results: Of 89 patients included in the development of the model, 39 were diagnosed with OMG. Based on our Bayesian network model, the following variables were the most useful predictors in descending order: edrophonium test, acetylcholine receptor (AChR) antibodies), single-fiber electromyogram (sfEMG), repetitive nerve stimulations (RNS) facial nerve, RNS accessory nerve, Besinger score, ice test, sustained upgaze test, dysarthria, dyspnea, dysphagia, diplopia, ptosis, age, and sex. The model was validated by determining the mean error rate and the area under the curve (AUC) by both 10-fold cross-validation and prediction on the retrospective USZ and UoT validation data consisting of 69 and 24 patients, respectively. Of all variables, edrophonium (sensitivity 94%, specificity 90%) and AChR antibody testing (sensitivity 85%, specificity 96%) showed the highest predictive value during validation with an AUC of 0.912 and 0.872, respectively. Incorporating more predictors reduced the predictive error in both validation data sets. Conclusions: Our prediction model serves as a basis to predict the OMG likelihood. It underwent successful internal and external validation and can be used to assist in clinical decision making.
2023
2021
- Student PaperModeling Choice in Co-DesignMarius FurterJul 2021
This report describes a method for modeling free and forced choice within Co-Design. In a free choice among a set, one has control over which option is selected, while in a forced choice one does not. Given a preorder P describing resources or functionalities, a free choice among a subset of P acts like a meet. Dually, a forced choice acts like a join. Moreover, the two types of choice distribute over one another. Based on this, we construct a universal model for choice on a preorder using the free completely distributive lattice ULP. Feasibility relations are then extended to these models. Along the way, we illustrate how to work within ULP and provide results that simplify calculations. The definitions presented here have been implemented in Haskell.
2020
- JGOThe targeted delivery of interleukin-12 to the carcinoembryonic antigen increases the intratumoral density of NK and CD8+ T cell in an immunocompetent mouse model of colorectal cancerEmanuele Puca, Caroline Schmitt-Koopmann, Marius Furter, and 5 more authorsJournal of Gastrointestinal Oncology, Aug 2020
The recent success achieved by immune checkpoint inhibitors in the field of immuno-oncology has been less evident for the treatment of metastatic colorectal cancer (mCRC) patients. To date, cancer immunotherapy has been efficacious only in few patients bearing high mutational burden (less than 25%) mCRCs. In this Communication, we report the generation of a novel antibody cytokine fusion protein (termed Sm3E-mIL12) targeting the CRC-associated carcinoembryonic antigen (CEA). The antibody moiety bound avidly to CEA when immobilized on solid supports, and selectively stained C51 tumor cells transfected with the antigen (C51-CEA). The cytokine payload retained full activity in vitro, as compared to the parental recombinant interleukin-12 (IL12). Ex vivo microscopic analyses revealed a homogenous distribution of Sm3E-mIL12 in the neoplastic mass upon intravenous administration. In vivo, Sm3E-mIL12 was well tolerated up to 180 µg per mouse. The targeted delivery of IL12 to CEA-expressing C51 carcinomas led to durable complete responses in 60% of the treated mice. The intratumoral density of immune effector cells was markedly increased after the third injection of Sm3E-mIL12, in keeping with the progressive regression of the neoplastic mass. The data suggest that a fully human analogue may be considered for the treatment of patients with mCRC.
2019
- Master ThesisEngineering Myoglobin Towards Halogenase ActivityMarius FurterETH Zurich, Mar 2019
Halogenated molecules serve as important synthetic intermediates and target compounds in medicine and agriculture. Enzymatic catalysis could allow for stereo- and regioselective installation of halogen moieties from simple halides in aqueous media. However, few readily engineerable enzymes are available for this task. Based on its structural similarity to chloroperoxidase, we envisioned the heme-enzyme myoglobin as being capable of halogenations. To achieve initial activity, a series of variants based on myoglobin H64V V68A (Mb*) were generated by combinatorial means. First, an aspartate or glutamate was installed in either position 43, 64 or 68. Second, the proximal porphyrin-binding histidine (H93) was replaced by a cysteine. Third, the iron-protoporphyrin IX cofactor was replaced by a synthesized manganese-protoporphyrin IX. The produced variants were screened for chlorination activity by monochlorodimedone (MCD) assay. Mb*A68D showed the highest initial rate of up to 30 per min, matching that of the chloroperoxidase, albeit only for up to 12 turnovers per active site. Variants bearing a cysteine as proximal ligand, and those containing the manganese porphyrin showed lower activities than less modified variants. Because of mechanistic similarity of halogenation and peroxidase reactions, both proceeding through compound I, peroxidase activity was measured by guaiacol and vanillic acid assays and compared with halogenase activity. The strong linear correlation of initial rates during MCD and vanillic acid assays establishes a link between the two reactivities that can be exploited by discovering potent halogenases among known peroxidases.