Protein Design Picks and Shovels
Nicholas Larus-Stone@nlarusstone
· 1
−4.5%
since Oct 3
The idea
I believe AI is changing how drugs get designed, not just discovered, and that shift plays out over many years, not quarters. Rather than guess which design engine wins, I own the picks and shovels every designed protein has to pass through: instrument makers and lab automation as the steady core, DNA synthesis and sequencing companies for more upside, plus a small slice of the compute behind the design models. I keep a cash buffer and spread it across many names rather than betting on one, and I'd rethink it if protein design stopped moving toward AI-driven methods.
What it owns
- The steady suppliers of machines and consumables that biotech labs buy year after year, collecting a toll across the entire shift regardless of which design approach wins. · Thermo Fisher ScientificSells nearly everything a biotech lab needs, so it collects revenue across the entire biotech spending shift regardless of which design engine wins., DanaherOwns the equipment and consumables used to develop and manufacture biologic drugs, earning more as AI-designed candidates flow through bioprocessing at higher volume., Agilent TechnologiesMakes the analytical instruments labs use to characterize and confirm designed molecules, benefiting as lab instrument spending in biopharma rises., BrukerSells machines that reveal protein structure, the exact property AI design models predict and labs must then confirm experimentally., WatersMakes mass spectrometers that confirm a designed protein matches its intended structure and composition., RevvitySells the screening tools and reagents used to test designed candidates at scale, so more AI-generated candidates drive more screening volume.38%
- Designed proteins must be sequenced and measured to confirm they work as predicted by the models. · IlluminaThe default platform for reading DNA sequences at scale, which labs need to generate the training data that feeds protein design models., QiagenSells the kits that prepare samples before sequencing or measurement, so it benefits automatically as lab activity around designed proteins grows., Pacific BiosciencesReads long stretches of DNA that short-read machines miss, providing the richer sequence data that trains better protein design models., 10x GenomicsEnables researchers to see what happens cell by cell, providing the fine-grained biological data that trains AI protein design models., QuanterixMeasures proteins at very low concentrations, the method labs use to validate whether a designed protein works as the model predicted.20%
- Once a protein is designed, these companies supply the equipment, consumables, and services to produce and test it at scale. · RepligenSells the filters and consumables used in every protein drug manufacturing cycle, earning on volume independent of which design platform wins., AvantorSupplies the everyday materials labs and biomanufacturers consume constantly, benefiting directly as overall lab and manufacturing activity grows., Charles River LaboratoriesRuns the safety and preclinical testing every drug candidate must pass before human trials, so more designed candidates means more testing volume.15%
- Every AI-designed protein begins as a synthetic DNA order that someone has to manufacture at scale. · Twist BioscienceThe largest supplier of made-to-order synthetic DNA, so it captures volume as labs order more DNA synthesis with each new designed protein., Ginkgo BioworksRuns the cell programming and build-and-test work that AI-first drug companies outsource instead of building their own wet labs., CodexisIt engineers the enzymes used to manufacture designed molecules, making it a key tool for producing therapeutics created through protein design.13%
- The AI models doing the protein designing all run on the same hardware platforms. · NVIDIAThe compute platform where protein design models run and train, so it captures hardware demand as biotech and pharma scale AI-driven drug design.6%
- Cash8%