ChromSword –
The unique Software for LC Method development of Large Molecules
Developing HPLC methods for large biomolecules (such as proteins, mAbs, peptides, and oligonucleotides) is a complex task because these molecules behave very differently from small molecules in chromatographic systems.
Working with the ChromSword package for large molecules, you can use two method development approaches: computer-assisted (offline) and automated (online). You can combine automated stationary and mobile phase screening with machine learning and fully automatic optimisation, or statistical design of experiments with retention and separation modelling.
Structural Complexity
Multiple Charge States
Complex Mixtures and Heterogeneity
Structural Complexity
Large biomolecules have:
- Multiple domains
- Disulfide bonds
- Post-translational modifications (PTMs)
- Flexible conformations
These structural features affect how the molecule interacts with the stationary phase, making retention and separation unpredictable.
Proteins and peptides have many ionizable groups – multiple charge states across pH ranges.
This means:
- Retention can change dramatically with small pH adjustments
These structural features affect how the molecule interacts with the stationary phase, making retention and separation unpredictable.
Complex Mixtures and Heterogeneity
Biomolecules often exist as
a mixture of variants:
- Glycoforms
- Deamidation, oxidation products
- Aggregates, fragments
Separating these closely related species is much more complicated than separating small-molecule impurities.
Statistical Design of Experiments (DoE) for HPLC method development of large biomolecules often fails or gives misleading results because the underlying assumptions of DoE do not hold for these complex analytes.
DoE assumes that responses change in predictable ways (linear or at least quadratic curvature).
But for large biomolecules, small changes in pH, temperature, ionic strength, or organic content can cause step changes rather than gradual ones.
Proteins may elute, stick, denature, or aggregate abruptly.
- This breaks DoE model assumptions and leads to poor predictive power.
Multiple interdependent effects occur simultaneously
For large molecules, HPLC retention is determined by a combination of:
- Conformation changes
- Unfolding/refolding
- Adsorption to surfaces
- Ionization state shifts
- Secondary interactions with metals and silica
Strong secondary interactions lead to unpredictable behavior
Biomolecules often interact with the silanol groups of a stationary phase; however, these interactions are not part of the design space and can unpredictably change retention.
DoE typically models:
- Main effects
- Two-factor interactions
- Quadratic curvature
But biomolecule separations commonly show:
- Three-factor and four-factor interactions
- Temperature × pH × gradient shape × ionic strength interactions
Standard DoE designs cannot capture these effects.
Heterogeneity of biomolecules adds hidden variables
Proteins exist as mixtures:
- Glycoforms
- Charge variants
- Fragments
- Aggregates
- Oxidation/deamidation species
Each species often responds differently to changes in chromatographic conditions. DoE assumes a single analyte responds consistently, but the same molecule can have many species with different chromatographic behaviors.
ChromSword’s automated method development algorithms for large biomolecules apply a mechanism-driven, adaptive, and iterative optimization engine rather than a predefined factor–response mode.
ChromSword uses real-time adaptive learning, not static design
Classical DoE requires you to predefine:
- which factors to vary
- the ranges
- the model form (usually linear or quadratic)
But large biomolecules exhibit nonlinear behavior, with abrupt changes in retention, peak shape, or selectivity.
ChromSword instead:
- Runs an experiment
- Reads the chromatogram
- Automatically adjusts conditions using data-driven learning, symbolic reasoning, and computational intelligence.
This closed-loop adaptive system handles nonlinear, unpredictable behavior far better than fixed DoE grids.
ChromSword optimizes nonlinear gradients, which are critical for large molecules
ChromSword automatically:
- designs step gradients
- applies curvature
- adjusts hold times
- distributes gradient segments intelligently
DoE cannot handle arbitrary gradient shapes. Humans take days or weeks to optimize them. ChromSword does it automatically.
ChromSword works directly with raw chromatographic behavior, not assumptions
DoE tries to correlate factors → responses (like Rs), but for large molecules:
- Rs is noisy
- Integration varies
- Peak positions shift in nonlinear ways, which leads to sharp maxima and minima in resolution plots
ChromSword instead:
- Evaluates peak movement and shape qualitatively
- Analyzes derivatives of chromatographic profiles
- Estimates peak capacity and selectivity without relying
- On noisy integrals
- Makes decisions based on chromatographic physics, not statistical assumptions
This makes it far more reliable for biomolecules.
Finally, ChromSword can run statistical design of experiments to identify well-optimized conditions for improving a method and to test its robustness.
Challenge in Large Molecules
- Nonlinear, unpredictable retention
- Complex peak families and shoulders
- Need for shaped gradients.
- Large, multidimensional parameter space
- Noisy integration
- Slow experimentation
Why ChromSword Succeeds
- Adaptive, iterative algorithm
- Advanced peak-recognition algorithms
- Automatic nonlinear gradient design
- Broad intelligent scanning
- Uses raw chromatographic behavior, not numerical Rs alone
- Fully automated exploration