Notes on bio-material development, inverse design, and the science behind Celleste.
Tensile strength is hard to estimate without a batch. We explain our Gaussian process method, inputs, and limits.
Read articleClosed-loop validation lets Celleste update lab-specific calibration from batch results, so predictions become more useful as a lab records more formulations and measurements.
ReadProtein-polymer composites sit between polymer science and protein biochemistry. This article examines the composition variables, interface tradeoffs, and literature limits that shape their design.
ReadComputational materials tools are often assumed to require more data than most labs have. This article explains where that concern is valid and how Bayesian methods handle sparse or incomplete datasets.
ReadA scaffold with 2% collagen by mass can support strong adhesion while reaching only tens to hundreds of pascals in Young's modulus. Here are the formulation choices behind that tradeoff.
ReadFormulation variables and bioreactor parameters are linked. A composition that holds up in static culture can fail under agitation or perfusion. We describe what Celleste models and what still requires lab judgment.
ReadCasting, organizing, and finishing lab-grown leather create demands unlike those for other biomaterials. The scaffold becomes the finished material, so its composition must support growth and long-term use.
ReadBiodegradation predictions have two uncertainty layers: a composition-model confidence interval and an environmental sensitivity flag. Together they explain why accurate in vitro results may not translate to implants.
ReadInverse design for bio-materials requires more than importing methods from inorganic chemistry. Smaller datasets, continuous formulations, process variables, and context dependence call for a careful methodological translation.
ReadTissue engineering scaffold formulation is harder to define than single-property optimization because several requirements must be satisfied at once. This article examines property interactions, constraint priorities, degradation, and transport.
ReadA four-week bio-material iteration may contain only five days of experimental work. The remaining 19 days go to waiting, logistics, analysis, and decisions that shape the next batch.
ReadA typical development program runs 20-100 batches covering less than 0.3% of even a modest composition space. This primer explains how generative design addresses that coverage gap in cultivated bio-material formulation.
ReadA researcher at a neighboring institution pointed out that a Celleste prediction for a collagen-hydroxyapatite formulation was off in a consistent direction. This post explains what that kind of proximity means in practice.
Read
