Pfentech

AI-Powered T-Cell Receptor Design

Pfentech generate affinity-matured, off-target screened TCRs
in under 2 hours - 1000x faster than in the laboratory

Meet the team

Isaac Pfaender

Co-Founder/CTO

Head of platform development, with Google DeepMind-funded research experience and multiple immune protein projects at Oxford University and UCL.

Prof. Benny Chain

CSO

Global leader in computational immunology with 40+ years of experience and a joint professorship in Immunology and Computer Science at UCL.

Bailey Costello

Co-Founder/COO

Head of operations, with a background in experimental physics, paired with extensive leadership experience through professional coaching.

What do we do?

TCR-T therapies have shown immense promise through clinical trials. But therapeutic discovery is slow, costly, and can result in potentially fatal consequences, when receptors bind to healthy tissue instead of the desired target.Pfentech was founded to fix all of this.Backed by UCL 'Innovation and Enterprise', we’ve built our proprietary, machine learning powered platform that designs, optimises, and safety-screens T-cell receptors in under 2 hours entirely in silico - removing the 3 months of laboratory trial and error usually required.Our rapid pipeline opens the door for world-first, completely bespoke treatments, not limited to just shared targets.

How we can help you:

If you already have a TCR candidate, we can improve its affinity and reduce its cross reactivity for the highest risk off-targets.And if you have a pMHC you want to target, we can create a tailored solution in just a few hours.

Our latest updates

Things are moving fast. See below for the latest Pfentech news!

1 hour.
87% accuracy.
Unseen data.

In-vitro results are in: Pfentech’s platform ranked unseen TCR mutants by affinity - see how our in silico predictions stood up to real-world laboratory testing.

See results

A fatal cross-reaction; flagged by Pfentech's platform

When a TCR unexpectedly bound to Titin, a protein found in cardiac muscle, 2 people died. Could our software have predicted this, and potentially saved lives?