10¹⁴+
unique barcodes (architecture ceiling)
The semiconductor industry made computation abundant through miniaturization. Molecular measurement has never had that moment. We bring it.
By shrinking the transistor and packing exponentially more onto each chip, every generation raised density and lowered the cost per bit, until storing and processing data at scale became routine. The measurement of molecular interactions has had no equivalent. Every drug discovery and diagnostic campaign asks which molecules interact, and how, yet today's best tools measure binding or activity for only a small fraction of what they screen. The record of measured interactions stays small and expensive to expand, starving machine learning of the data it needs, most acutely for covalent and macrocyclic chemotypes, patient-serum antibody responses, and molecular diagnostics.
Hyperplexing
Identity and activity, read in one pass, at scale.
Hyperplexing means reading identity and activity together, in a single pass, across a population of thousands to millions of beads at once, not one sample at a time, but the whole library simultaneously. Beads embedded with random microdisks create an optical fingerprint that uniquely encodes every bead. As that population streams through, each bead's identity and result stay linked to it, turning what used to be one-at-a-time reads into one connected, population-scale dataset.
Encode the library
Each library member (small molecule, macrocycle, peptide, oligo, or antibody) is coupled to its own uniquely barcoded SpectraGel™ bead. Laser particles distribute randomly during microfluidic droplet formation, specifying one of trillions of barcodes.
Pool & add sample
The whole library goes into one well with the target or sample: everything reacts together in a single shared volume. No robotics. No plate handling.
Bind & wash
Actives capture or convert their target; unbound material washes away and a reporter labels the result. On-bead affinity or off-bead biochemical and cell-based activity: both modes supported.
Read & decode
The whole population (thousands to millions of beads) is interrogated in one hyperplexed pass, by infrared spectral flow cytometry or a planar microwell array. Each bead's spectrum gives identity; its visible-channel reporter gives the result, for every bead, including the non-interacting majority.
Map
Barcodes decode to members, signals to interactions: out comes the complete interaction matrix, every compound linked to its measured activity.
Peer-reviewed foundations
The laser-particle science the platform builds on, published in leading journals.
Proof of concept
Demonstrated. Not just modeled.
Three results prove the architecture works today: identity, uniqueness, and completeness, all measured on real beads.
10⁶
Beads decoded per set
Across million-bead sets, each bead's spectral barcode decodes to its library member, with duplicate barcodes held below 0.01% between independent sets.
0%
Duplicate barcodes at k ≥ 5
Duplicate barcodes fall to 0.02% at four particles per bead and reach zero at five or more. A single additional lasing peak renders a bead essentially unique at billion-bead scale.
100%
Of the population read
Every bead is interrogated and decoded, actives and non-actives alike, so weak effectors and true negatives are retained in the dataset rather than lost.
The data ML has been missing
Every connection, from one read.
AI can predict a protein's structure, but it can't observe interactions at scale. The limit isn't the algorithms: it's that almost nothing in the training record says which molecules were tried and did nothing.
Every bead carries both its identity and its result, so one campaign resolves into a map: which compounds acted on which targets, at what strength, and which did not.
Why SpectralTx
The empty corner of the encoding map.
Every screening method must link a molecule's identity to its activity: completely, at scale, with a consistent and relevant assay. Every incumbent trades something to do it. SpELs can be designed to interrogate new chemical space: optical encoding liberates library design and scale, beads provide a scaffold for binding or compound release into nanoliter volume compartments, and in-situ decoding matches library ID with activity across every member in one pass.
/ UNCONSTRAINED CHEMISTRY
Beyond the aqueous-only regime.
SpectraGel's inert encoding system permits a broad chemistry tool-kit needed to generate sp³-rich, macrocyclic, covalent, and natural-product-like chemical space: structures underrepresented in most compound/data repositories, yet offering ample opportunity to discover novel activity and specificity in emerging therapeutic areas.
/ PRECISE OPTICAL DECODING
Sub-nanometer lines. Calibrated confidence.
Fluorophore-encoded bead systems are typically limited to a few hundred codes by overlapping emission. Razor-sharp laser-particle lines (<1 nm FWHM, 1150–1650 nm infrared) decode to the exact member with calibrated, per-bead confidence, delivering clean identity calls across millions of beads at once, with zero optical interference to the assay readout.
/ COMPLETE DATA · DEEP LEARNING
The negatives are the gold.
Because every bead is IDd regardless of the assay performance, each campaign yields a comprehensive, compound-to-target interaction matrix. Agonists, partial inhibitors, weak effectors, inactives: all retained, every replicate IDd, every assay signal connected. That structured, library-wide data is the fuel deep-learning needs, providing agency to classify and train a model without the empirical method conflating a selection or sorting survivor bias.
We give predictive biology what selection-based screening can't: the full population, identity and activity together, in one pooled read.
The Spectral Therapeutics mission
Why it scales
More combinations, not more colors.
Plexity comes from combining a small set of laser-particle spectral lines into distinct signatures, not from making more colors. A few peaks per bead, drawn from a library of ≈1000 distinguishable infrared wavelengths, generate an address space far larger than any fluorescent or DNA-oligo barcode. Today we have >10⁶ codes. Raising loading to four particles reaches >10⁸, and the architecture supports >10¹⁴, enough to catalog every protein in every proteome on Earth.
Address space at each complexity level
>10⁶
codes today
2.35% barcode replicate
>10⁸
near-term
<0.02% barcode replicate
C(1000,k) combinatorial space, enough to catalog any proteome, compound library, or diagnostic assay without duplicate barcodes.
Where SpELs fits alongside HTS & DEL
What if you could read them all?
1 tube
one pooled reaction runs the whole library: multiple screens per day, where conventional HTS is slow and cost-prohibitive at library scale*
Read, not sequenced
identity and activity decoded optically, in the same pass, where DEL requires a separate sequencing run to recover which survivors bound*
Filling the library-wide data gap
Selection-based approaches like DNA-encoded libraries (DEL) recover identity by sequencing the survivors of a binding selection, so most of a library yields no usable data and negatives stay unmeasured, and their chemistry is largely limited to mild aqueous conditions. SpELs is designed to complement these methods: it reads every bead, retains the negatives, and generates data across the sp³-rich and covalent chemistries where today's predictive models have the least to learn from.
* Representative internal estimates. Actual figures vary with library size and assay configuration.
About us
The team behind SpectralTx.
Spectral Therapeutics is built by a founding team working at the intersection of semiconductor photonics, microfluidics, and drug discovery: the disciplines spectral encoding has to bring together.
Co-founder & CEO
Illumina · 1859 Inc. · Yale · NUS
Seventeen years building miniaturized quantitative high-throughput experimentation across multi-omics and drug discovery; co-founded 1859 Inc., a deep-tech company (85+ FTEs) that industrialized one-bead-one-compound DNA-encoded library technology.
Co-founder & CTO
1859 Inc. · Plexium Inc. · Scripps Research
Leads bead technology and encoding: instrumentation, microfluidics, and bead manufacturing; Fifteen years in drug-discovery miniaturization. Scientific co-founder OBOC-DEL - trained in Brian Paegel's laboratory at Scripps Research, where functional bead-based DEL screening was developed, and served as Director of Screening at 1859 Inc., running campaigns against more than fifteen targets.
Co-founder & CSO
Abbott · Neurocrine · BioBlocks · 1859 Inc. · Caltech
Leads chemistry. A medicinal chemist with over three decades of leadership and an extensive patent portfolio, specializing in high-throughput solid-phase chemistry, DNA-encoded library design, and hit-to-lead optimization.
Founding team & CIO
1859 Inc. · Plexium Inc. · Scripps Research
Leads data integration: decoding, data architecture, and the interaction data the platform generates. Co-founded Plexium and 1859 Inc. Scientific co-founder OBOC-DEL - trained in Brian Paegel's laboratory at Scripps Research.
Supported by
Grateful for the backing behind bold science.
National Science Foundation
Grant #2528478
National Institutes of Health
Grant #1R43TR005303-01A1
Partnerships
We partner selectively.
SpectralTx works with a small number of partners on significant programs: collaborating closely to build spectral encoding around the questions that matter most to your pipeline, rather than selling material off a shelf. If complete, library-wide interaction data could change how your team makes decisions, we'd like to talk.
Discuss a collaboration →
Read the science
hello@spectraltx.com