Posters & Pitches

NanoTech manufacturer for precise rapid tests of Food ingredients – for food safety and QM

Author: Robert B Brunner

Innovative solutions for human and veterinary medicine, environmental analysis and the food industry.

Our precise rapid tests are used for the detection of analytes such as pesticides, toxins, cytostatics and antibiotics.
The products of NanoTech @ Help-Doctor can be stored at room temperature from 2 °C to 30 °C.
This is ideal for an uncomplicated on-site application.

USP:
Faster detection of analytes such as pesticides, toxins, cytostatics and antibiotics.
within in food and pets.
Precise results within 3 to 15 minutes. No laboratory testing necessary!

Market / Target groups:
Labs, veterinary/ veterinaries, pet owners, gastronomy/cooks/chiefs, food industry

Diagnostics (Biomedical), Diagnostics (Medical Technology) Laboratory Technology: quanti-SERS or qSERS, the next generation of rapid tests.
Our qSERS are used for the on-site determination and absolute quantification of substances (we call them analytes). An analyte can be a drug (e.g. cytostatics and antibiotics) in the blood of patients or contamination of antibiotics / mycotoxin in food.
Ultra-stable nanomaterials: The main customers of nanomaterials are various diagnostic, cosmetic and chemicals industry. The nanomaterial is also used by the manufacturing of fuel cells and solar cells.

Nanodiagnostics and POCT (point of care tests) #lab #IVD
In addition, our innovative method can also be used for other laboratories. e.g. for quick and easy monitoring of toxins in food, feed or in water (environmental technology).

Method:
Surface-enhanced Raman spectroscopy or surface-enhanced Raman scattering (SERS) https://en.wikipedia.org/wiki/Surface-enhanced_Raman_spectroscopy e.g., see the Schematic illustration of synthetic SERS nanotags (a) and quantitative detection of E. coli O157:H7 using the SERS-based lateral flow assay (b) https://www.researchgate.net/figure/Schematic-illustration-of-synthetic-SERS-nanotags-a-and-quantitative-detection-of-E_fig1_344242033

Diseases:
• The Feline Leukaemia Virus (FeLV) is a retrovirus which can cause cancer. 
• Our lyme test is a rapid immunochromatographic test of animals after infection with Borrelia.
• A reliable antigen detection of parvoviruses in the faces  of cats and dogs. Parvoviruses cause the disease parvovirosis and are primarily transmitted via faces or stomach contents (vomit).

2.
We’re at Help-Dortor.de are a “MedTech development shop” specializing in IT integration, AI-powered imaging, endoscopy … Interested in Horizon 2000 projects and others?
We’ve developed a digital system for laboratory processes that integrates with LIMS, offers AI support, and simplifies data collection in research and laboratory settings for MDR compliance. This has been developed as a prototype/MVP. We are currently seeking additional partners to submit a research proposal.

From Gut Microbiome Profiles to Explainable Brain-Health Interpretation: A Partner-Facing Platform Integrating Microbial Taxa, Metabolites, Host Pathways, and European Nutrition Translation

Author: Minkyung Park

Directly translating gut microbiome composition into health conclusions remains biologically and commercially limited because taxonomic abundance alone does not adequately explain functional output, host signaling, or intervention relevance. I present an explainable gut-brain interpretation platform designed to bridge this gap by connecting microbiome readouts to metabolite-linked and host-pathway-aware interpretation layers, and then translating these outputs into partner-ready reports and localized nutrition guidance.

This gap matters because the microbiome market is clearly moving beyond basic taxonomic reports. Commercial platforms increasingly combine microbiome testing with dietary recommendations, wellness framing, and brain- or mood-facing insights, yet recent expert consensus also emphasizes that the clinical usefulness of microbiome testing remains limited and that interpretation requires caution. The current market therefore has a translation problem: consumer-facing outputs are expanding faster than biologically explicit, mechanistically traceable interpretation systems.

My platform is designed to fill that gap by linking microbial taxa to metabolites, host mediators, receptors or pathways, and brain-related outcome domains in an explicit evidence structure. Instead of generating opaque summary scores, the system produces traceable interpretation chains that show why a specific microbial pattern may matter, where the mechanistic bridge is stronger or weaker, and how confidence should be communicated. To reduce overclaiming, the workflow includes a human-gated literature review queue, explicit separation between approved knowledge and draft evidence, and a production architecture in which newly approved papers can expand the knowledge graph without automatically changing scoring logic or recommendations. 

On the product side, the platform supports a public demo layer, a partner-facing report layer, and a cohort-oriented workflow for follow-up interpretation over time. A further differentiator is a European nutrition translation layer, developed to adapt microbiome-linked food guidance to realistic regional dietary contexts instead of relying on generic one-size-fits-all recommendation logic. I argue that the next generation of gut-brain microbiome applications should move beyond composition-only outputs toward systems that are biologically grounded | mechanistically traceable | partner-usable.

USP39-SpliceTACs: First-in-Class Allosteric Degraders Targeting Spliceosome Fidelity in Cancer

Author: PD Dr. Xinlai Cheng

- USP39-SpliceTACs convert a previously undruggable spliceosome fidelity regulator into a  degradable cancer vulnerability, offering a new route to selectively collapse tumor proteostasis.

Cancer cells operate under chronic proteotoxic and transcriptional stress. To survive, they depend on adaptive quality-control mechanisms that preserve RNA processing, protein homeostasis, and cellular fitness. We identified USP39 as a cancer-enabling spliceosome fidelity regulator that helps tumor cells maintain proteostasis under stress (Prieto-Garcia et al. 2024 Science). Although USP39 is annotated withindeubiquitinase famility, it’s catalytically inactive, acts as a scaffold protein, and considered aschemically undruggable by small molecule inhibitor. 

Using our well-established AI-assisted drug discovery platform and DEL screening (> 600 M compounds), we are developing a new class of USP39 allosteric degraders, including conventional PROTAC-inspired molecules and proteasome-targeting degraders (ProGraders), to selectively destabilize USP39 and trigger lethal proteotoxic stress in cancer cells(Schäfer et al. Angewandte Chemie, 2026). Our preliminary data show successful generation of USP39-SpliceTACs, selectively and robustly degrading USP39 in cancer cells and a liver cancer mouse model. USP39 loss is expected to impair spliceosome fidelity, disrupt cancer-cell stress adaptation, and induce tumor cell death through proteome imbalance.

This project combines mechanistic cancer biology with translational drug discovery. The therapeutic  concept is especially relevant for high-unmet-need malignancies such as hepatocellular carcinoma and other stress-adapted tumors where broad proteostasis modulation has been clinically limited by toxicity. By targeting a cancer-specific vulnerability upstream of proteostasis collapse, USP39 degradation may provide a more selective and druggable intervention point.

Beyond a single asset, the project establishes a platform logic for targeted degradation of non-classical oncology targets. Our approach expands the degrader toolbox from ligand-driven PROTACs toward alternative proteasome-targeting strategies, enabling access to proteins that are difficult to inhibit but biologically decisive for tumor survival. The current development goals are hit-to-lead optimization, validation of cancer selectivity, biomarker definition, and positioning for preclinical translation.

At BayOConnect 2026, we aim to connect with biotech partners, investors, translational mentors, and drug discovery experts to acceleratedevelopment of USP39 degraders toward a first-in-class therapeutic program.

AGRODRUG SYNTHESIS: CHEMICAL STRATEGIES FOR THE OPTIMIZATION OF SMALL MOLECULES FOR DROUGHT RESISTANCE

Author: Dr. Isaac Garcia

During project DROUGHDRUGS “Drugs for thirsty crops”, we worked on the design and synthesis of new molecules able to activate the abscisic acid (ABA) receptor in plants, responsible for the resistance of plants against drought.

Part of the strategy involved the derivatization of already known structures, such as Sulfobactin (N-benzyl-1,4-dimethyl-2-oxo-1,2-dihydroquinoline-6-sulfonamide).[1] So as to generate comprehensive structure-activity relationships (SAR), different substituents (R1, R2, Ar) were introduced in the structure. More importantly, we carried out detailed modifications on the linker between the quinolinone and phenyl ring of Sulfobactin, a moiety known to be of great relevance for their potency as ABA agonist. [1-3]. 

We describe some of the chemistry developed ad-hoc during this project to complete the different series of target compounds.

References
[1] J. Lozano-Juste, J. et al., Sci. Adv. 2023, 9 (10), eade9948. 
[2] F. C. Okamoto et al. Proc. Natl. Acad. Sci. U.S.A. 2013, 110 (29), 12132-12137.
[3] M. Bono et al. Molecular Plant, 2025; 18, 1526-1548.

RetroSCOPE technology effectively detects mRNA and non-coding RNA from up to 10 year old FFPE preserved samples 

Author: Dr. Callum Coupland

The formalin-fixed paraffin embedded (FFPE) method has been used in clinics for over 130 years but presents challenges for genetic analysis, including crosslinking of DNA and RNA, fragmentation, chemical modifications, and RNA degradation. These limitations, alongside the vast number of archived samples, necessitate improved approaches for accurate RNA profiling. Probe-based hybridization methods for FFPE analysis are limited by predefined targets, restricting broader transcriptome exploration. Singleron’s RetroSCOPE technology employs random oligo priming to capture all nuclear RNA, enabling detection of variants, gene fusions, isoforms, allele-specific expression, gene copy number, and non-coding RNAs, including microRNAs. Its full-length capture strategy minimizes 3′ and 5′ bias seen in other methods. Nuclei are extracted from FFPE tissue, and reverse transcription introduces barcodes to cDNA, followed by partitioning on the Singleron SCOPE-chip for capture and downstream library preparation. The method demonstrates robust transcriptome coverage from partially degraded RNA and has been validated on clinical and decade-old human and mouse samples. It enables reproducible nuclei resolution and gene expression analysis, including clustering based on non-coding RNA. RetroSCOPE technology provides a powerful approach for unlocking transcriptomic insights from archived FFPE samples.

On a Paradigm Shift in CAR-T Cell Manufacturing

Authors: Michaela Wagner, Theresa Kagerbauer, Thibaud Rivière

The clinical success of CAR-T cell therapy in B-cell-associated malignancies provides clear evidence that cell therapy has the potential to play a key role in the treatment of a wide range of diseases and has encouraged many researchers to identify potential target molecules which could accelerate the advancement of cell therapy beyond the CD19 antigen. However, even access to proven CD19-targeted cell therapies remains blocked for many patients in need due to logistical challenges and the time-consuming and costly conventional CAR-T manufacturing processes, which slows the progress of cell therapy regardless of the target. With the aim of eliminating logistical challenges in a time- and cost-saving “vein-to-vein” process, while maintaining the safety profile of controllable cell manufacturing, we report here on the development of a process that enables the generation of functional CAR-T cells directly from whole blood within a few hours.

In a fully automated, two-hour cell-hold step, we isolate a clinically relevant amount [1] of highly pure T cells directly from 200 ml of whole blood. Our non-magnetic, column-based cell selection matrix is coated with low-aOinity and fully reversible selection reagents [2] and enables a rapid and cost-eOective CD3-positive selection, as these CD3-positively selected T cells exhibit all characteristics (full TCR expression, no pre-activation or intracellular responses) comparable to CD3-untouched T cells. Building on proven cell therapies targeting the CD19 antigen,we have developed a genetic modification approach generating T cells that express the CAR temporarily for up to four days. Unlike conventional modification approaches, purified T cells are transfected with CAR mRNA in their resting state, thereby preserving the initial memory subset compositionand the non-activated phenotype in the final CAR-T cell product. Using various mRNA transfection methods, including electroporation or lipid nanoparticles (LNPs), the CAR mRNA is therefore introduced instantly after T cell isolation, with complete CAR surface expression detectable 4–5 hours after genetic modification. These CAR-T cells are fully functional, which we confirmed by specific lysis of target cells and a significant cytokine production in response to recognition of the CD19 antigen in vitro.

In summary, we have developed a manufacturing process that is suitable for use at the bedside, as it enables automated generation of CAR-T cells directly from whole blood within a few hours. The transient CAR activity oOers the unique advantage of repeated administration and has the potential to regulate immune imbalances or abnormal immune responses as needed, while simultaneously reducing the risk of long-term toxicity or persistent immunosuppression, suggesting a promising application in the treatment of autoimmune diseases [3].

References
(1) Rotte A, Frigault MJ, Ansari A, Gliner B, Heery C, Shah B. Dose-response correlation for CAR-T cells: a systematic review of clinical 
studies. J Immunother Cancer. 2022 Dec;10(12): e005678. doi: 10.1136/jitc-2022-005678. PMID: 36549782; PMCID: PMC9791395.
(2) Radisch, S., Poltorak, M. P., Wagner, M., Cletiu, V., Radisch, C., Treise, I., Pann, S., Weigt, A., Artner, S., Dreher, S., ... Germeroth, L. 
(2022). Next generation automated traceless cell chromatography platform for GMP-compliant cell isolation and activation. Scientific 
Reports, 12(1), Article 6572. doi.org/10.1038/s41598-022-10320-x
(3) Wu D, Xu-Monette ZY, Zhou J, Yang K, Wang X, Fan Y and Young KH (2025) CAR T-cell therapy in autoimmune diseases: a promising 
frontier on the horizon. Front. Immunol. 16:1613878. doi: 10.3389/fimmu.2025.161387

Carbacitabine (CAB): A Next-Generation Hypomethylating Agent Overcoming Current Limitations in Cancer Therapy

Author: Dr. Matthias Heiß

In elderly and medically non-fit patients with acute myeloid leukemia (AML) or myelodysplastic syndrome (MDS), treatment options are severely limited. Intensive chemotherapy, targeted agents, and immunotherapies are frequently not feasible in this population due to their toxicity burden. Hypomethylating agents (HMAs) such as azacitidine (AZA) and decitabine (DEC) therefore represent the standard of care in this high risk setting, yet their clinical activity remains limited by susceptibility to hydrolytic and enzymatic cleavage.

By directly targeting this instability, we developed Carbacitabine (CAB) as a next-generation HMA: a single structural modification, replacing the ring oxygen of the ribose moiety with a CH₂ group, generates the carbocyclic analog of DEC. This on a molecular level minimal change has far-reaching consequences: CAB acquires resistance to hydrolysis while retaining DNMT-trapping as its primary mechanism. In a dose-escalation toxicity study in mice, blood counts (leukocytes, erythrocytes, platelets) remained within the physiological range at CAB doses more than 100-fold above those causing hematotoxicity with DEC. This dramatically reduced toxicity is particularly relevant for highrisk patients with limited physiological reserves, as it preserves tolerability across treatment cycles and, by permitting higher therapeutic doses, contributes to superior anti-leukemic efficacy. In patientderived xenograft (PDX) AML models, CAB reduced tumor burden more effectively than DEC and/or AZA at equitoxic doses, with complete remissions observed in a subset of animals. CAB further induced apoptosis in multi-resistant patient-derived AML cells unresponsive to DEC and/or AZA, indicating activity against known resistance mechanisms and raising the prospect of effective treatment for patients who have exhausted all other options.

CAB is currently in late-stage preclinical development within the group of Prof. Dr. Thomas Carell at LMU Munich, with a defined roadmap to the clinic which was already reviewed and accepted by theregulatory authority. Remaining milestones include GLP toxicology studies, GMP manufacturing, and finalization of the clinical study protocol, with Phase I/IIa trial initiation targeted for 2028. To advance CAB into the clinic, a spin-off company is planned. Right now, the project is called EpiCure and is actively seeking capital from both public funding programs and private investors.

We present an enzymatic linkage approach that enriches the current pool of chemical, antibody-based and enzymatic methods, impacting research and biomanufacturing capabilities

Author: Dr. Adrian Fuchs

Next-Level bioconjugation of proteins to ligands
Protein conjugation is foundational across modern bioscience, powering an ecosystem from protein detection and immobilization to drug payload delivery. We present an enzymatic linkage approach that enriches the current pool of chemical, antibody-based
and enzymatic methods, impacting research and biomanufacturing capabilities. 

Today's methods, chemical (NHS, maleimide), click chemistry, enzymatic (Sortase A, TGase), and split-domain systems (SpyTag, DnaE) are seriously limited with low yields, side reactions, suboptimal conditions, and complex workflows. In R&D, that impacts detection (sensitivity, quantification), immobilization, and yield, tag-based and with antibodies. The goal is a defined and stable bond between protein and conjugate/s, best for multi-purpose use. Overall, current methods create heterogeneous conjugates and/or side products. For ADC developers, that results in lower therapeutic indexes and increased downstream purification.

Connectase, discovered by Dr. Adrian Fuchs et al. at the Max Planck Institute for Biology, Tübingen, solves limitations at their core. Its unique reaction mechanism eliminates side reactions entirely. The 13–19 amino acid recognition sequence ensures convincingly high specificity without off-targets. With a hydrolysis-resistant amide bond it achieves 98–100% conjugation yield, when combined with proline aminopeptidase (PAP). That works with both N- and C-termini. Connectase’s catalytic efficiency is surprisingly robust, performing well also under mild physiological conditions and in otherwise challenging crude extracts. The setup approaches the maximum yield even within minutes. Switching ligands is an option. 

We plan to make a set of “Research Tools” broadly available to the scientific community, starting in mid-2027. For Diagnostics and Pharmaceutical development, here in the field of advancing Antibody-Drug Conjugates, we embrace the application of our research 
and inventions. 

The Proof of Concept is published (PNAS 2021; Nature Communications 2023; Scientific Reports 2024; eLife 2025) and two patent families are filed. EXIST Forschungstransfer Phase I provides non-dilutive funding through end of 2027, underpinning our path to commercialization. We appreciate discussions, review, and input. We look for partnerships with additional R&D labs, industry players, and investors.

Application of FCCS in the Development of Bispecific Antibodies

Author: Prof. Dr. Uwe Jacob

Fluorescence Cross-Correlation Spectroscopy (FCCS) is a versatile platform technology for investigating molecular interactions under physiological conditions. The method has been successfully applied to small-molecule drugs, peptides, antibodies, membrane proteins, and nucleic acids to determine equilibrium binding constants and kinetic parameters. 

FCCS assays can be performed in complex biological environments, including crude cell lysates, body fluids, living cells, blood plasma, and tissue homogenates. The technology provides a highly precise and information-rich readout, enabling the determination of molecular concentrations, molecular size, binding states, and complex half-life. 

In this work, we focus on the characterization of a bispecific antibody by assessing target occupancy in cells and tumor biopsy samples. FCCS enables quantitative analysis of target engagement and molecular interactions directly in biologically relevant systems, supporting translational and preclinical drug development.

2NA FISH: A Novel Spatial Transcriptomics Platform for Spatial RNA Biomarker Detection and Clinical Practice

Author: Dr. Christina Port

Background
Spatial transcriptomics has become indispensable for understanding the tumor microenvironment (TME), immune cell infiltration and therapy resistance - yet adoption in clinical routine remains blocked by six-figure instrument investments, multi-day workflows and analysis pipelines requiring specialized bioinformatics teams. As a result, the molecular insights that should be informing therapy decisions today are confined to research settings. A diagnostically deployable solution represents a significant unmet need at the intersection of precision oncology and routine 
diagnostics.

Methods 
2NA FISH is an integrated spatial RNA diagnostics platform combining a proprietary DNA-nanotechnology-based multiplexed RNA FISH assay with the 2NAlyzer, an AIpowered web application for tissue segmentation and per-cell quantification. The assay currently simultaneously measures 7 RNA biomarkers on a single FFPE tissue slide - extensible to panels of up to 50 targets including microRNAs, a class poorly addressed by leading spatial transcriptomics platforms. Critically, 2NA FISH runs on standard fluorescence microscopes already present in pathology labs: no microfluidics, no dedicated instrumentation, and no specialized training required for interpretation.

Results 
In a bladder cancer cohort, 2NA FISH achieved high-resolution ERBB2 biomarker detection on FFPE samples with strong concordance to clinical IHC scoring - and delivered superior signal quality compared with established spatial transcriptomics competitors. MicroRNAs were detected at high resolution on cell culture samples, opening biomarker categories largely inaccessible to current platforms. The 2NAlyzer software provides a traceable workflow with built-in error checks and an AI-based segmentation model that achieves robust performance without requiring membrane staining - removing one of the main analytical bottlenecks in spatial RNA quantification. Together, the platform delivers a per-slide cost an order of magnitude below large-scale spatial transcriptomics systems.

Conclusion
2NA FISH is positioned as the bridge between research-grade spatial transcriptomics and routine clinical diagnostics. As the method runs on standard pathology equipment and requires no specialized training for analysis, it enables integration of spatial RNA biomarkers into clinical routine. Given the accelerating demand for spatial RNA biomarkers in patient stratification - particularly across complex solid tumors where the TME drives therapy response - 2NA FISH is built for partnerships with pharma, diagnostics players and pathology networks looking to operationalize spatial biology at clinical scale.

T-TOP - We bring TCR-T-cell therapies TO every Patient

Author: Dr. Susanne Wilde

Every day, nearly 27,000 people die from cancer because their immune systems fail to recognize and eliminate tumor cells. T-cell receptor (TCR)-T cell therapies offer a promising new approach by reprogramming a patient's own T cells to specifically detect and kill cancer. However, current TCR-T therapies face key limitations: tu mors can escape treatment through heterogeneity, only a few effective TCRs exist for different patient populations, and manufacturing in centralized GMP facilities is slow (3-6 weeks), complex, and costly.  T-TOP addresses these challenges through an integrated platform that combines artificial intelligence (Al), experimental validation, and rapid decentralized manufacturing. Our unique "Lab-in-the-Loop" approach features a high-throughput natural TCR discovery platform in combination with G.O.A.T. (Generator of Al-designed TCRs), a proprietary generative Protein Language Model that designs and validates multiple patient-tailored TCRs for tumor-specific mutations and antigens. Al-designed TCRs are rapidly validated in T-TOP's wet-lab platform, with the resulting data being continuously fed back into the model to improve prediction accuracy, efficacy, and safety. This iterative optimization bridges computational design with real-world biological performance, enabling the development of clinically relevant TCRs with greater confidence than purely in silico approaches. By generating diverse sets of complementary TCRs rather than relying on a single receptor, T-TOP reduces the risk of tumor immune escape while expanding treatment accessibility to all patients. To overcome the scalability and cost barriers, T-TOP combines its Al-driven design engine with an EXiVO manufacturing process developed in collaboration with a local partner, enabling patient-specific TCR-T cells to be generated and delivered directly at the patient's bedside within < 24 hours and to reduce treatment costs of < $40,000 per patient.  T-TOP's master product will target HPV-associated cancers leveraging a lead TCR that has already demonstrated clinical efficacy with tumor size reductions > 30% in 4/7 patients (Li et al., ESMO 2025). HPV-associated cancers and an estimated $2.2 billion market alone, will pave the way for T -TOP' s expansion into personalized cancer therapies for other indications. T-TOP aims to transform TCR-T therapy from a highly specialized treatment into an affordable global standard, making life-saving T-cell therapies accessible to every patient.

Accelerating hit-to-lead conversion in drug discovery

Author: Dr. Jean-Philippe Sobczak

At tilibit we revolutionize hit-to-lead conversion in drug discovery by addressing the two main challenges of in-vitro functional characterization and throughput with our single-molecule chip technology. 

Existing in-vitro analytical devices for molecular characterization cannot reproduce the spatial molecular configurations required for biological function, leading to false positives with high failure rates in later stages. Our technology enables controlled presentation of drug targets at biologically relevant distances and geometries on the nanometer scale, allowing 
predictive screening and selection of novel biologics candidates without expensive in-vivo testing. We can selectively characterize cis- and trans-binding, receptor recruitment, and other complex functions, identifying true therapeutic candidates instead of artifacts. By 
detecting individual molecules, we can also characterize sample heterogeneity and ultrastable binders that are challenging or impossible to characterize with current tools. These unique single-molecule kinetic fingerprints provide a novel way of IP diGerentiation, 
providing value even beyond the direct R&D impact.

Large language model agents cannot reliably copy biological sequences: sequence copy infidelity as a new risk to life-sciences automation

Author: Nicholas Lee

AI agents are being rapidly deployed into biotech and pharmaceutical workflows, where they autonomously process biological sequences for bioinformatic pipelines, patent database searches, and functional analysis. The drive for operational efficiency has accelerated a parallel trend toward Small Language Models (SLMs), which offer a cost-effective foundation for automation compared to expensive flagship models. The optimism behind this trend, however, rests on an assumption that has not been tested: that these models can faithfully reproduce the sequences they are meant to act on.

For drug development, the utility of an LLM-based interface is contingent on its ability to handle biological sequence data with zero margin for error. In this study we identify a new risk associated with agentic AI for the life sciences, which we term "sequence copy infidelity": errors in biological sequences introduced by the LLM itself, either when formulating tool calls, including those issued through standardized protocols such as the Model Context Protocol (MCP), where the model must generate the sequence token by token through its own output stream, or when displaying processed sequences back to the scientist. 

We benchmarked this phenomenon and found that error rates increase sharply with decreasing model size. However, even flagship models such as GPT-5.2 achieved less than 90% per-base accuracy when reproducing low-complexity sequences of only 1000 nucleotides. The SLMs favored for cost-effective agentic automation collapsed to near-zero fidelity, a performance entirely unacceptable for any purpose in the life sciences.

Ironically, LLMs struggle to reproduce the low-complexity repeat regions where biological DNA replication and transcription itself is most error-prone.

Our results expose a critical vulnerability in how transformers process sequence data. Reliable bioinformatic agents require either fundamental improvements in copy fidelity or a careful redesign of the tool-use paradigm. Until this is addressed, general-purpose LLM agents should not be deployed on tasks where sequence integrity is required.

DEVELOPMENT OF NOVEL DRUG CANDIDATES FOR SELECTIVE INHIBITION OF Bcl-2 PROTEIN–PROTEIN INTERACTIONS

Author: Dr. Jairo Paz

The Bcl-2 family of proteins is central to the regulation of apoptosis, and their overexpression has been linked to certain types of lymphoma and carcinoma [1], as well as resistance to conventional antitumor treatments. In particular, the interactions between pro- and anti-apoptotic Bcl-2 family proteins control the integrity of the outer mitochondrial membrane. The aim of this project was to develop a set of novel anti-cancer drug candidates that function by selectively inhibiting such protein–protein interactions (PPI) [2].

In collaboration with HQL Pharmaceuticals (Israel), a virtual library of potential candidates was designed and further refined following ADME, stability, structural complexity, and patentability criteria. The virtual hits were clustered into families and selected compounds were chosen for 
synthesis. The activity of the compounds, that is, their capability to inhibit Bcl-2 PPI, was evaluated by Surface Plasmon Resonance (SPR) measurements. After the initial hit-finding stage, the structures were further optimized (hit-to-lead) to improve the activity down to the 
nanomolar range.

This work was carried out within the projects ONCOGALFARMA and NEOGALFARM, funded by the Galician Innovation Agency (GAIN) and FEDER through the CONECTA-PEME Programme.

References:
[1] Otake Y., Soundararajan S., Sengupta T.K., Kio E.A., Smith J.C., Pineda-Roman M., Stuart R.K., Spicer 
E.K., Fernandes D.J., Blood, 2007, 1:109, 3069.
[2] Mullard, A. Nat. Rev. Drug Discov, 2012, 11, 173

Spatial analysis of physical activity and ambient air pollution in Kuwait: a cross-sectional study

Author: Ralf-D. Schroers

Air pollution may be associated with lesser physical activity (PA), a major risk factor for cardiometabolic disease. Such an analysis has not been conducted for Kuwait, a major petroleum producing nation with some of the highest rates of diabetes in the world. This study aimed to assess the associations between spatiotemporally expressed ambient air pollutant concentrations and PA in Kuwait. A cross-sectional study was conducted from 2011 to 2014, involving 2529 adults ≥ 18 years. Participants were classified as physically active if they exceeded 600 Metabolic equivalents (MET) minutes per week. Monthly mean air pollutant concentrations (hydrogen sulphide, nitric oxide, nitrogen dioxide, nitric oxide, nitrogen oxides, ozone, and PM10) were spatially interpolated to local residential areas using data from local monitoring stations. Random effects logistic regression analysis was conducted to assess the associations between PA and air pollutants (an alpha level of 0.05 was used for all statistical tests). Participants were predominantly male (63%) with mean body mass index 29.3 kg/m². 52% were physically active. The overall air pollutant score was inversely associated with PA (adjusted odds ratio: 0.87, 95% CI 0.79–0.95). A 0.01 mean increase (ppm) in air pollutant concentration was associated with proportionately lower odds of being physically active for hydrogen sulphide (10%), nitric oxide (14%), nitrogen dioxide (16%), and nitrogen oxides (14%). In contrast, a 0.01 increase (ppm) in ozone concentra)on was associated with 15% greater odds of being physically ac)ve. Our findings highlight for Kuwait a high level of insufficient PA and a previously undocumented negative association between PA and air pollution, particularly nitrogen-based pollutants and hydrogen sulphide. Research is required to evaluate how air pollution may inhibit PA and whether reductions in emissions could increase PA in Kuwait.

Automated Patient-to-Trial Matching Teaser: An NLP-based platform that extracts trial-relevant data from unstructured clinical records and matches patients to eligibility criteria, deployable on-premise or in the cloud and GDPR-compliant by design.

Author: Mahi Butt

Background.
Patient recruitment is the single biggest cause of clinical trial delays. Around 80% of trials miss their enrollment target and timeline, and delays incur substantial cost. The underlying cause is structural: eligibility screening remains a manual process, in which clinical staff review unstructured records against complex inclusion and exclusion criteria, costing hours of chart review per patient. This is slow, costly, and does not scale. It also concentrates trial activity at a small number of large academic centers, limiting patient access and representative real-world data, particularly for novel therapies. 

Approach.
Rosentres automates eligibility prescreening. The platform ingests structured data and unstructured free-text physician letters and uses NLP to extract and normalize relevant clinical parameters. Extraction is performed by fine-tuned medical language models, initially for German and designed to extend to further languages and specialties. Trial eligibility criteria are decomposed into individual inclusion and exclusion conditions, and candidates are retrieved via lexical and semantic search. Each candidate is assessed criterion by criterion, and every match is accompanied by a traceable, human-readable justification, presented to clinicians through a dedicated review dashboard. The system is built on a FHIR-based data model to support interoperability with existing hospital infrastructure. Data protection is integral to the design: patient data is pseudonymized, enabling a GDPR-compliant cloud deployment, with on-premise operation where local data residency is required.

Outlook.
Validation is currently conducted on a representative clinical dataset, enabling criterion-level evaluation against a physician-defined reference standard. Building on this, prospective validation with clinical research partners and trial sites is planned to assess matching performance under real recruitment conditions, with the objective of reducing manual chart review from hours to minutes while preserving data sovereignty. Initial exchanges with clinical partners corroborate both the severity of the recruitment bottleneck and the demand for a compliant solution. By lowering the operational barrier to trial participation, Rosentres has the potential to broaden recruitment beyond specialized centers and is presented at BayOConnect 2026 to connect with clinical sites, CROs, and industry partners across the European trial ecosystem. 

rosentres.de

Accelerating Drug Discovery with Rapid Mouse Model Generation

Author: Dr. Benedikt Wefers

The generation of genetically engineered mouse models remains one of the most consequential - and often most frustrating - steps in preclinical drug discovery. Delays compound, timelines slip, and the downstream costs of waiting for a model that may still not behave as expected are rarely trivial. Addressing this bottleneck requires more than incremental improvements to established workflows; it calls for a rethink of how models are designed, produced, and selected in the first place.
Ozgene has developed two complementary technologies that directly target the core constraints of speed and scientific rigour. goGermline™ eliminates the inefficiency of conventional chimera breeding by producing only ES cell–derived animals with confirmed germline transmission, compressing the path to first experimental cohorts without sacrificing genetic quality. OzBIG enables large genomic replacements using bacterial artificial chromosomes, allowing full humanisation of complex loci - including regulatory elements and intronic sequences - in a single step. The translational implications are significant: rather than approximating human biology, these models can replicate it at the sequence level, with direct consequences for target validation and therapeutic relevance.
Selecting the right model, however, is as important as building it well. Building a custom model from scratch is not always the fastest or most pragmatic route. Depending on the research question, timeline, and IP considerations, purchasing a commercially available model or in-licensing an academically derived line may better serve the project. We will briefly discuss a framework for navigating this decision - and where the limits of each approach tend to emerge in practice.
Whether you are at the stage of target identification, lead optimisation, or reconsidering an existing model that is not performing as expected, we welcome the conversation.
 

Beyond the Bench: Nucleate Germany's Blueprint for Translational Biotech

Author: Đesika Kolarić

Nucleate is the largest global trainee-led, non-profit organization dedicated to empowering the next generation of biotech leaders by bridging the gap between scientific discovery and venture creation. Since its founding as a Core chapter in 2023, Nucleate Germany has grown into a multi-city network spanning 4 regions across major German research hubs, including Munich, Berlin, Heidelberg and North Rhine-Westphalia, engaging PhD students, postdocs, and early-career scientists who are exploring entrepreneurship as a career path. 

At the heart of Nucleate Germany's impact is the Activator program, an equity-free curriculum that guides interdisciplinary founding teams from early venture concepts to investor-ready companies. To date, 26 founding teams have completed the program, several of which have gone on to found their companies, demonstrating the program's role in de-risking the earliest stages of biotech venture formation. 

Beyond the Activator, Nucleate Germany runs an engaging calendar of community events — panels, workshops, and networking sessions — connecting 1400+ scientists with mentors, investors, and industry leaders each year. Nucleate’s flagship Demo day, BioInnovation Forum, and region-specific gatherings like Munich's "Bavaria meets Biotech" alone have brought together 500+ attendees including founders, VCs, and pharma representatives, fostering cross-sector dialogue that is otherwise rare for trainees, and facilitating knowledge exchange, interdisciplinary collaboration, and career development for emerging talent in life sciences.

This growth reflects a broader shift: German academic institutions are increasingly recognizing translational entrepreneurship as a viable and valuable career trajectory for scientists. Nucleate Germany's model - community-first, peer-led, and embedded within the academic ecosystem - offers a scalable template for activating Europe's research talent toward company-building, helping address the well-documented gap between Europe's scientific output and its commercialization rate.

Antibody Alternatives for Infectious Diseases: A Discovery and Engineering Platform for Protein-Based Binders

Author: PhD Peter Braun, Julius Frey

Antibodies remain the standard for molecular recognition in infectious disease diagnostics and in therapy. Their discovery, however, relies on animal immunization and mammalian cell culture. A workflow with long development cycles, complex and costly production, limited stability requiring cold-chains, and inherent difficulty distinguishing closely related pathogen species. At the same time, the rising threat of antimicrobial resistance (AMR) and the persistent risk posed by biothreat agents create growing demand for faster, more specific, and broadly deployable binders for diagnostic and therapeutic use. At the Fraunhofer ITMP–IIP Penzberg Lab, we are developing a discovery and engineering platform for proteins that specifically bind pathogens and serve as leads for diagnostic and (combined with effectors) therapeutic development using two approaches: A) De novo binder design generates novel binding proteins against clearly defined target structures, such as viral structural proteins or bacterial outer membrane proteins. AI-based design tools (e.g. BindCraft) are followed by expression in E.coli as fluorescent or enzymatic fusion proteins, biophysical characterization (including binding kinetics and thermal stability), and in vitro validation.

(B) Utilizing naturally evolved phage receptor binding proteins (RBPs) that have co-evolved with their bacterial hosts over millennia and recognize complex bacterial surface structures. From publicly available phage genome sequences, we use in silico structure prediction and RBP-specific filters to narrow several hundred candidate proteins per genome to a tractable shortlist, which is then recombinantly expressed and tested for binding against the target bacterium (under BSL-3 conditions where required). The platform is co-developed with its application across a growing portfolio of pathogen-specific projects. Targets include multi-drug resistant clinical pathogens such as Klebsiella pneumoniae and Acinetobacter baumannii, as well as highly pathogenic biothreat agents such as Bacillus anthracis and Yersinia pestis.

Urinary tract infections are the most common bacterial infection worldwide, yet standard diagnostics still take 48–72 hours, forcing clinicians to prescribe broad-spectrum antibiotics empirically. This practice is a major driver of antimicrobial resistance - one of the greatest threats to global public health.

BugSense is developing a multilayer paper-based in-vitro diagnostic that simultaneously delivers species identification (94% accuracy), bacterial load quantification, and antimicrobial resistance profiling within 4–12 hours at the point of care, with no laboratory infrastructure required. The test is designed for use in GP practices, emergency departments, and care homes. The technology builds on a proprietary combination of selective chromogenic media, lateral flow filtration, and automated image analysis. An IVDR-compliant intended use has been defined; clinical validation is underway.

BugSense is funded by the Bavarian M4 Award and EXIST Forschungstransfer (combined >€1.5M) and is currently seeking strategic industry partners and seed investors to accelerate clinical certification and market entry. 

Website: bugsensedx.com

Speaker: Sarah Wali, CEO

Drug development remains highly inefficient, as many promising drug candidates fail in clinical phases due to the limited predictive power of conventional preclinical models. New approach methodologies (NAMs), including perfused vascular and tumor models, offer a promising path towards more human-relevant testing and are pushed by regulatory developments such as the FDA Modernization Act 2.0. However, their broader adoption is still limited by complicated handling, insufficient standardization, and a lack of automation.
CELLenger overcomes these barriers with a cell culture platform that brings physiologically relevant, continuous perfusion into the standardized and easy-to-handle well-plate format. It seamlessly integrates biologically meaningful 2D, 3D, organoid and organ-on-chip models into existing laboratory workflows and infrastructure – enabling automation and parallelization of next-generation cell culture models.
CELLenger is a pre-seed startup of the University of Augsburg that is currently supported by the Bavarian Validation Funding Program and has recently received a funding recommendation from the EXIST Research Transfer jury. With this support, the team is accelerating product validation and preparing the platform for a successful market launch.

Website: https://www.linkedin.com/company/cellenger

Speaker: David Wörle, Founder

Coldex is developing a first‑in‑class antiviral nasal spray that provides immediate, broad‑spectrum early‑treatment efficacy and protection against respiratory viral infections. Respiratory viruses remain a major global health burden, with limited options that act fast enough to prevent disease progression. Coldex introduces a proactive, patient‑friendly solution: a locally delivered RNA-LNP formulation that activates the body’s natural RIG‑I antiviral pathway directly at the site of infection: the nasal epithelium. This mechanism triggers rapid innate immune defense within minutes, reducing viral replication early and offering protection against diverse pathogens, including influenza, RSV, rhinovirus, and SARS‑CoV‑2. Our optimized RIG‑I ligand and clinically validated lipid delivery system create a potent, stable, and safe therapeutic platform. Coldex bridges the gap between prevention and treatment: a single spray can both block infection before it starts and reduce viral load when administered shortly after exposure. With strong IP, a clear regulatory path, and a high‑value entry market of 350M high‑risk individuals, Coldex aims to establish the world’s first rapid, real‑world antiviral defense standard.

Speaker: Christine Wuebben

ENDOLEASE is a first-in-class implantable platform technology for controlled intra-arterial drug delivery. The system is designed to be placed minimally invasively into an arterial vessel, where it releases therapeutics directly into the bloodstream supplying the downstream target tissue via its capillary system. This approach enables high local drug exposure at the site of action while aiming to reduce systemic toxicity and improve therapeutic efficacy. The technology addresses a major unmet need across several indications in which systemic treatment is limited by side effects, insufficient local concentrations, or repeated treatment cycles. Potential applications include interventional cardiology, oncology, neurology, and transplantation medicine. 

Unlike conventional drug-eluting stents, ENDOLEASE is not primarily designed to treat the vessel wall, but to deliver drugs through the arterial blood flow to the target organ or tissue region. The platform combines catheter-based implantation, biodegradable materials, and controlled drug-release concepts. Its modular design allows adaptation to different drugs, release profiles, and clinical indications. ENDOLEASE is currently being developed by an interdisciplinary team from the University Hospital Würzburg, combining clinical expertise, biofabrication, materials science, and translational entrepreneurship. 

With its precision-delivery concept, ENDOLEASE aims to establish a new therapeutic paradigm: local efficacy, reduced systemic burden, and scalable application across high-need medical indications.

Website: https://endolease.de 

Speaker: Anna Fleischer, Project lead

Website: https://www.flashtd.com/ 

Speaker: Karim Tabiti, CEO

Nucleaf addresses a critical gap in modern agriculture: the need for sustainable, adaptable, and effective alternatives to conventional chemical pesticides. Increasing regulatory pressure, rising resistance among pests, and climate-driven shifts in pathogen dynamics demand new solutions that combine efficacy with environmental compatibility. Our approach leverages RNA-based technologies as a programmable platform to develop highly specific crop protection agents that can be rapidly designed and iteratively optimized for different targets.

In the near term, we generate revenue through RNA design, synthesis, and validation services for agrochemical and seed companies. This not only enables early market entry but also builds proprietary datasets on RNA efficacy, stability, and delivery in plant systems. These data strengthen our platform and guide the development of our own target-specific candidates. As we validate RNA agents against relevant agricultural pests, we pursue co-development and licensing partnerships, contributing candidates and performance data to accelerate downstream development and regulatory approval. 

By integrating scalable synthesis, high-throughput design, and rapid biological testing, our platform shortens development cycles and improves targeting precision over time. Nucleaf positions itself as a data-driven platform partner for next-generation crop protection in Europe.

Website: www.nucleaf.de

Speaker: Timo Schlemmer, Founder, & Florian Stokom, Founder