Molecule Design

Well-designed molecules are at the core of many innovative biosolutions, in life science and in other industries. Our proprietary platform supports targeted and efficient, yet also creative in silico design, augmentation, simulation and validation of such new molecules and molecule types.

Such molecules include biological antibodies, vaccines and enzymes, with targeted functions and features. Yet, our  molecule design functions are not purely conditioned on proteins. They also cover lipids, glycans, nucleic acids and biological or biocompatible polymers and conjugates, including binders, linkers and payloads. For enzymes, such conjugates include cofactors and immobilizing matrices and scaffolds.

Our platform also enables de novo design of such molecules. It also allows augmentation, re- and multi-purposing, combination or merely characterization of existing proprietary and/or published third-party molecules. Design functions range from molecule generation, structure determination, conjugation, docking and perturbation to ADMET-profiling and approval-filtering. Design extensions include polypharmacology mapping and molecule re- and multipurposing.

What molecules do we design?


Past and ongoing projects with different partners include:

  • Designing novel antibody-drug-conjugate (ADC) binding to a heavily glycosylated antigen, linked to a polymer carrier, for diagnostic and therapeutic purposes in oncology.
  • Designing novel nanobodies (VHH) for therapeutic strategies in oncology, penetrating blood-brain-barrier and targeting brain tumor cells with cytostatic payload.
  • Simulating and visualizing specified polymer carrier with polyethylene glycol dimethacrylate (PEG-DMA) units, including radioisotopes in DOTA chelators.
  • Modifying alginate lyase enzyme for accelerated cleaving of the mannuronic acid (M) and guluronic acid (G) polymer in pelagic brown algae Sargassum fluitans III.
  • Simulating fatty acid photodecarboxylase (FAP) photoenzyme, powered by light and heat, targeting biofuel production from biomass in photobioreactors (PBR).
  • Augmenting several biocatalysts, including MHETase metalloenzymes e.g. with metal-switching for more efficient degradation of polyester plastics and textiles in domestic waste.
What are molecule design functions & features of our platform?


Our proprietary platform allows in silico molecule design for specific medical and industrial application. Key functions and features for molecule design include:

  1. Generation
    1. Generative AI methods:
      • Conditional diffusion model;
      • Variational autoencoder (VAE);
      • Generative adversarial network (GAN);
      • Graph neural network (GNN) generator.
    2. Evolutionary algorithm (on SMILES, graph structures, latent vectors, etc.)
      .
  2. Structure Determination
    1. Conditional on biomolecule class & type (proteins, glycans, lipids, biocompatible polymers, e.g. PEG, PLA, PCL);
    2. For proteins: Apply homology & AI-based algorithms;
    3. For all: Apply rule-, force field- & molecular dynamics-based algorithms.
      .
  3. Conjugation
    1. For antibody drug conjugation:
      • Target conjugation & cleavage strategy;
      • Enhance binder(s) (e.g. for click chemistry);
      • Complement payload (e.g. by NHS ester, DBCO, maleimide);
      • Optimize conjugation (e.g. for efficiency, specificity, binder folding preservation);
      • Cross-validate results (e.g. molecular size, weight, energy); Simulate cleavage (e.g. by enzyme assumed available).
    2. For enzyme immobilization:
      • Target conjugation type (e.g. covalent, affinity, adsorption, entrapment);
      • Ascertain enzyme stability, activity, compatibility;
      • Modify support (e.g. membrane, polymer, e.g. by NHS ester, aldehyde, epoxy);
      • Ascertain conjugation parameters (time, pH, temperature, ionic strength);
      • Cross-validate results (e.g. load efficiency, thermal & pH stability, catalysis cycles).
        .
  4. Docking
    1. Determine…
      • …binding pose, epitope & paratope;
      • …binding affinity & function;
      • …association & dissociation;
      • …buried surface area;
      • …∆ energy (shape, desolvation, vdWaals, electrostatics, ionic bonds, H-bonds).
    2. Consider…
      • …conformational change of binding molecule;
      • …enthalpic and entropic parameters;
      • …different force field models;
      • …competing binding poses, epitopes & paratopes;
      • …other ligands covering this pose.
    3. Observe…
      • …signaling;
      • …cleavage;
      • …activation;
      • …inhibition.
        .
  5. Perturbation
    1. Trigger
      • …structure change (e,g, bond rotation, residue mutation, monomer addition);
      • …ambient change (e.g. pH, temperature, solvent);
      • …interaction (e.g. binding partner, ligand concentration);
      • …signaling (i.e. propagating disturbance through interaction network).
    2. Monitor
      • …atomic coordinates;
      • …bonding topology;
      • …torsion angles & rotamers;
      • …electron density & partial charges;
      • …free energy, enthalpy & entropy.
    3. Consider
      • …different free energy models;
      • …different molecular dynamics parameters.
        .
  6. Toxicity Prediction
    1. Observe…
      • …chemical structure & functional groups;
      • …absorption, accumulation, solubility;
      • …partial charges, electron density, reactivity indices;
      • …binding affinity to toxicity targets;
      • …activation of toxicity pathways.
    2. Mitigate…
      • …by removing toxicophores (with QSAR guidance);
      • …by substituting fragments (with safer analogs);
      • …by tuning charge distribution (reducing reactive metabolites);
      • …by modifying metabolic hotspots;
      • …by reducing accumulation in sensitive organs (e.g. liver & heart);
      • …by modifying binding affinity to hERG (avoiding cardiotoxicity).
        .
  7. ADME Profiling
    1. Determine…
      • …solubility;
      • …permeability;
      • …metabolic stability;
      • …CYP enzyme inhibition;
      • …mutagenicity;
      • …bioavailability;
      • …clearance.
    2. Mitigate…
      • …limited solubility by adding or tuning ionizable groups, reducing lipophilicity;
      • …limited permeability by adding hydrophobic groups, reducing polar surface area, achieving passive diffusion, and reducing entropic penalty crossing membranes;
      • …limited biodistribution by reducing plasma protein binding, reducing aromatic surface, and adding polarity in non-critical regions;
      • …excessive metabolism by finding and blocking metabolic soft spots, adding small steric shields, and reducing easily oxidized motifs; …limited clearance by adding polar groups, and introducing metabolic soft spots.
        .
  8. Approval Filtering
    1. Design-pass: Using classic rules like Lipinski, Veber, Ghose, PAINS;
    2. ADMET-suitability: Using rules from points (6) & (7) above;
    3. Approval-likeliness: Proprietary VAE.
      .
  9. Design Extension
    1. Map polypharmacology: identify, predict and validate all molecular targets the compound interacts with, map and review related therapeutic and off‑target effects;
    2. Re- & multi-purpose molecule: predict secondary targets, off-targets, beneficial multi-target profiles, and repurpose for new indications.
What are key challenges addressed by molecule design?


This question triggers a vast variety of comments, leaning towards very different answers. In our experience to date, the major challenges are revolving around molecular complexity in practice, in particular regarding conjugation and polymerization.

Complexity

For complex and information-dense molecules, such as large glycans, multi-domain proteins, or DNA, the computational burden for an in silico platform may be significant.

  • Some enzymes, such as fatty acid synthases, are large, multi-domain proteins, performing multi-step catalysis with unstable intermediates, pivotal in many metabolic pathways;
  • Large, glycosylated tumor antigens such as CA-125 in ovarian, pancreatic and lung tumors reach sizes exceeding 20’000 residues, with masses of 2-3MDa, due to repeat sections;
  • Human ingestion of microplastic is involving large polymer chains, such as polyethylene terephthalate (PET), with an efficient degradation pathway yet to be developed.

An in silico platform must master molecules of significant size and complexity to meet growing in silico expectations and needs. Generative modeling, perturbation and/or simulation of such molecules, e.g. for docking or cleavage, requires efficient representation, a suitable and scalable platform architecture, and generous hardware.

Conjugates

The ideal molecule solving your tasks may not be a pure protein, glycan, lipid or nucleic acid. Although each molecule type has its standard applications, nature tends to conjugate molecules for additional stability and functional  properties.

  • Your ideal antibody and its target antigen may be naturally glycosylated.
  • This may significantly condition epitope and paratope, impact binding affinities etc.
  • Your ideal enzymes may have a metal cofactor, with crucial oxidative functions.
  • Such metalloenzymes may cleave even non-biological molecules, such as polyethylene.
  • Your ideal drug molecule’s permeability, solubility and bioavailability may benefit
  • from linked esters. These may augment your drug’s entire ADME pathway.

Hence, in silico tools and techniques should master properties and features of natural and artificial conjugates to realize their benefits in practice. Puristic protein design alone may not offer the molecular solution space needed for most practice applications.

Polymers

Many in silico applications involve polymers in a driving or supportive, intended or unintended role:

  • The ideal molecule solving your tasks may have been designed as monomer, but oligomerize in practice, by nature. This may significantly impact the functionality and bioavailability of your molecule. The likelihood and extend of oligomerization and the oligomer structure and function can typically be determined in silico.
  • On the other hand, your ideal molecule may have been designed as oligomer or polymer. This is the case for nanobody-polymer-conjugates, with the polymer carrying cytotoxic drugs or radioisotopes. In silico tasks include development of optimally sized polymers with high internal drug compatibility, external bioavailability and suitable release kinetics.
  • Furthermore, your ideal molecule may be targeting polymers. This is the case for many enzymes cleaving polymers back to oligo- and monomers. Such as cellulase cleaving cellulose to glucose or cutinases cleaving polyethylene to ethylene glycole and terephthalic acid. Finally, your ideal molecule, such as the aforementioned enzymes, may need to be immobilized and/or stabilized by a polymer serving as scaffold or forming a polymer-enzyme-composite.

While deterministic or stochastic polymer generation and drug loading in silico is well understood, it requires substantial computational resources for creation and perturbation due to molecule size and combinatorial complexity.

How do we differentiate?


As a focused in silico solution and service provider, we are offering dedicated collaborations, exclusive partnerships and joint ventures. We differentiate from other providers with such offerings in terms of application breadth, quantitative scrutiny and solution scope. These three features may be critical success factors for your biosolution development and/or productization project…

  1. Breadth of Application
    We are serving multiple industrial segments, including life science, chemicals, energy, utilities, food, textiles and construction material. The versatility and scalability of our in silico technology is anchored in the joint biological and chemical principles of the diverse industry applications.

    Resulting benefits:
    • Our in silico platform allows cross-industry benchmarking and validation, re-purposing of bioproducts, and extended data utilization and machine learning.
    • Our in silico platform allows efficient configuration to new and completed applications, without the burden of complexity-driving and time-consuming customization.
    • Our in silico platform promotes out-of-the-box thinking and guides cross-discipline innovation on a shared scientific basis.
      .
  2. Quantitative Scrutiny
    Data acquisition in vivo or in vitro, through lab experiments, typically comes at a high price. In silico technologies can compute insightful parameters and indicators with lower time and resource spending. Yet, type and accuracy of these parameters strongly depends on the underlying in silico technology. Some AI approaches generate only heuristic data. Stochastic approaches yield probabilities. Not all scores meet decision makers’ expectations.

    Our approach:
    • We are offering advanced metrics supporting design decisions at various levels. These metrics are scientifically grounded, explainable, understandable, justifiable and comparable.
    • For these metrics, we are offering proprietary and/or syndicated benchmarks, internal experience and external reference values.
    • We are monitoring consistency between metrics and benchmarks, securing plausibility of the overall design parameters, and engineering trust in the solution, conclusion and insights.
      .
  3. Scope of Solution
    Computational development of biosolutions does not end with proven molecule manufacturability, affinity and function. In silico technology takes biosolution development on a journey much further and broader,  leading from early target projection via molecule design and pathway engineering to a computational validation, trial and/or approval simulation and finally to IP protection.

    We differentiate by supporting a fairly large portion of this in silico capability spectrum:
    • Project Targets: Mine biological data sets, screen patent databases and identify targets. Prioritize targets based on scientific insight, technological possibilities and commercial opportunity.
    • Design Molecules: Generate hit and lead molecules for a given purpose. Determine their natural structure(s), ascertain their function(s) (QSAR) and preview properties (ADME) in their target environment.
    • Engineer Pathways: Determine how molecules behave in their biological network. Consider molecule features (e.g. solubility, permeability, stability, toxicity) and interaction (e.g. activation, inhibition, catalysis) along their metabolic or regulatory, human or industrial pathway. Adjust the pathway by modifying molecule(s), ambient conditions etc.
    • Simulate Trials (currently not implemented on our platform): Model cells, tissue, organs, patients and cohort permutations in a virtual clinical trial environment. Predict outcomes and adjust original model parameters to secure clinical and approval success.
    • Protect IP: Secure the ultimate commercial success of your innovation by utilizing in silico technology for patent formulation. This may even be extended to peparing regulatory registration, certification or approval, hazard classification, etc. 

▷▷▷ Contact us to learn more?