DocumentRX-8158
Issued
Read4 min

Tesorai Raises $5.3M Seed to Mine Raw Proteomics Spectra with AI

Ex-Verily founders raised $5.3M seed ($8M total) for Tesorai, whose AI reads raw mass spec spectra, finds 12–68% more peptides and reprocessed five brain studies in three weeks.

TAG K-1215 · 782 words on the filing

Former Verily Leaders Build AI Platform to Get More from Proteomics Data
Former Verily Leaders Build AI Platform to Get More from Proteomics Dataaxevaper / pixabay

Summary of findings

  • Tesorai raised $5.3M in seed funding this week, bringing total funding to $8M.
  • Tesorai Search identified 12–68% more unique peptides than MaxQuant, FragPipe, PEAKS and Proteome Discoverer in a Journal of Molecular Biology paper.
  • Tesorai Chat exceeded 92% accuracy on a 94-case proteomics benchmark.
  • An Alzheimer's reanalysis of five brain studies recovered unreported p-tau217 in four datasets and took three weeks.
  • Token-based pricing replaces annual licensing; transcriptomics and genomics are next on the roadmap.

Tesorai has closed $5.3 million in seed funding, bringing total funding for the proteomics AI startup to $8 million and financing an expansion from mass spectrometry data into transcriptomics and genomics.

The round drew Perplexity Fund, IQT, South Central Ventures, New Stack Ventures, NextGen Venture Partners, f7 Ventures, VitalStage Ventures and ZAKA. The company, founded in 2024, has already signed contracts with biopharmaceutical and research organizations, with named collaborators including teams at MD Anderson Cancer Center, New York University and the Michael J. Fox Foundation.

The three co-founders met at Verily, Alphabet's life science company. CEO Melissa Miao led business initiatives and strategic partnerships; CTO Peter Cimermancic, PhD, led artificial intelligence teams; CSO Charlie Kim, PhD, launched and led Verily's molecular science program. "That's what inspired us to leave and build our own team specifically focused on building AI models to maximize the insights that you could extract from those data sets," Miao told GEN.

What does the platform actually do?

The core is a proprietary AI model that reads full mass spectrometry spectra from new lab experiments, public repositories and company archives. Its deep-learning search engine, Tesorai Search, is described in a Journal of Molecular Biology paper. The company reports that Tesorai Search "consistently identifies more unique peptides compared to established search engines like MaxQuant, FragPipe (+MSBooster), PEAKS, and Proteome Discoverer," with "average increases of 12–68%." The engine works out most of the 20 to 40 settings that standard tools require users to set manually, and shows "strong performance even on data not encountered during training."

Two further components round out the stack:

  • Tesorai Trove catalogs curated public proteomics datasets from human tissue and structures study metadata into searchable tables usable by humans and AI. Existing tools "only query and mine" condensed literature datasets, Cimermancic noted; Trove lets scientists "mine raw data directly."
  • Tesorai Chat runs quality checks, statistics and charts from conversational prompts. On a 94-case proteomics benchmark shared earlier this year, it reached over 92% accuracy, ahead of tools including Phylo and GPT-5.4.

Training strategy matters here. Instead of fitting models on the fly — which Cimermancic said risks overfitting — Tesorai pre-trains on heterogeneous public-repository data so the model generalizes across experimental conditions and instrument types, and therefore "doesn't necessarily have any biases." Validation is cross-vendor: models trained on one manufacturer's instruments are tested on another's. Academic partners have also built experimental evaluations, such as detecting single-gene knockouts in cell lines. Tesorai received its first platform patent last year and has invested in infrastructure to reprocess massive repositories at scale, because "software wasn't keeping up" with advanced instruments, Miao said.

The Alzheimer's proof point

Kim pointed to a recent application that shows what reanalysis of existing raw data can deliver. Tesorai tested a newly published multiprotein plasma panel for distinguishing early- from late-stage Alzheimer's disease — described in a JAMA Neurology paper this year — against data from five existing brain studies.

An initial read showed the datasets appeared to be missing p-tau217, an established plasma biomarker of amyloid-beta pathology. "Because phosphorylated tau 217 (p-tau217) and amyloid beta 1-40 (Aβ40) were not reported for any studies, raw spectra were reprocessed to quantify these proteoforms," the scientists wrote. p-tau217 turned out to be present in four of the five original datasets but absent from the reported protein summaries. Across cohorts, the plasma-derived panel distinguished early from late-stage disease, and the analysis "provided a mechanistic explanation for why these proteins were changing in the blood."

The entire analysis took three weeks, with no new study run. "There can be major accelerations in being able to validate and answer questions like this," Kim said of combining multiple existing datasets.

What does this change for buyers?

Pricing departs from the annual-license norm in proteomics software: customers pay through a token system. Academic teams receive more generous free credits, Miao said, while biopharma engagements are often structured as service contracts that include working within a client's cloud environment or building cross-department analysis sharing. "But for the most part, we're token-based pricing," she said.

The roadmap extends beyond proteins. The company will take on transcriptomics data next, then genomics, with the stated goal of "being able to combine different layers of biological data to assemble a mechanistic understanding," Kim said. "We see applications across all drug development or most drug development stages," Miao added, from target discovery through biomarker discovery to later development — plus a push into agrolife sciences.

More signal from every sample, faster: that is the pitch, and $8 million now backs it.

via sciencedirect.com (Original)

Share this article:

More from Olivia Hart

Olivia Hart

Show full bio

Senior reporter covering money, funding and market shifts around global pharmaceutical industry at PharmInsider.

22 articles

Linked filings

  1. E-9238
  2. K-4629
  3. E-6328
  4. K-5499
  5. K-8037

« Previous filing