Deep O igi docki g softwa e a d molecula dy amics simulatio s pi poi ted the lead compou d f om a 17-compou d lib a y
Computatio al esults alig ed with labo ato y expe ime tatio esults led by esea che s at Sta fo d a d MD A de so Ca ce Ce te a d published i the jou al Cell
SOUTH SAN FRANCISCO, Calif., July 20, 2026 (GLOBE NEWSWIRE) — Deep O igi today a ou ced that its computatio al platfo m co t ibuted to a pee – eviewed study published i Cell i which the compa y’s docki g a d molecula dy amics simulatio s successfully atio alized the esea che s’ wo k ide tifyi g a pote t compou d to switch o cell-death p og ams i diffuse la ge B cell lymphoma (DLBCL).
The study, “A Bivale t Molecula Glue Li ki g Lysi e Acetylt a sfe ases to O coge e-i duced Cell Death,” was led by esea che s at Sta fo d U ive sity a d the U ive sity of Texas MD A de so Ca ce Ce te . It desc ibes the developme t of KAT-TCIPs (lysi e acetylt a sfe ase t a sc iptio al/epige etic chemical i duce s of p oximity), which act as molecula glues that i duce a te a y complex betwee the ge e-activati g e zymes p300/CBP a d the ca ce d ive B cell lymphoma 6 (BCL6). The i te actio is of i te est to ca ce esea che s because it suppo ts pote t cell killi g. Molecula glues, a type of CIP facilitati g the assembly of u elated p otei s, is a fast-eme gi g d ug class.
The esea che s at Deep O igi used p op ieta y docki g softwa e, molecula dy amics (MD) simulatio s, a d qua tum mecha ical calculatio s to assess a lib a y of 17 compou ds. Fi st, the te a y complexes we e ge e ated by docki g each compou d at the p otei -p otei i te face a d selecti g poses with the best docki g sco es. The , Deep O igi ’s molecula dy amics e gi e was used to u 300- a oseco d simulatio s fo each te a y complex to efi e the i itial pose a d ge e ate a e semble of st uctu es. Fi ally, the qua tum mecha ical calculatio s we e utilized to calculate li ke st ai e e gy fo each compou d based o the obtai ed st uctu al e sembles. The team a ked the compou ds by the e e getic cost that they i cu whe fo mi g the te a y complex.
The compou d that i cu ed the least e e getic cost killed lymphoma cells at sub- a omola co ce t atio s (IC50 0.80 M).
The computatio al esults alig ed with labo ato y tests. I a mouse xe og aft model of the disease, the selected compou d led to complete o ea -complete tumo clea a ce, a d i immu ized mice it depleted ge mi al ce te B cells, which exp ess high levels of BCL6 a d se ve as a model fo ce tai lymphomas, with o ove t o ga toxicity.
“Deep O igi ’s computatio al simulatio s flagged the compou d that would kill ca ce cells most effectively – matchi g esults achieved at the be ch,” said Ga egi Papoia , Ph.D., co-fou de a d chief scie tific office of Deep O igi . “This is a p edictivity goal of i silico d ug discove y – to dete mi e the ca didates most likely to achieve desi ed esults p io to wet lab expe ime tatio . At Deep O igi , we’ve made g eat st ides i computatio al p edictio by combi i g the igo of physics-based simulatio with the speed a d scale of AI, outpe fo mi g i dust y be chma ks. This was a mea i gful eal-wo ld test of ou systems.”
TCIPs act th ough a gai -of-fu ctio mecha ism that appea s to activate cell death p og ams much mo e effectively tha BCL6 deg ade s a d i hibito s.
“We desig ed KAT-TCIPs to ec uit the lysi e acetylt a sfe ases p300/CBP to BCL6 to activate ep essed cell death ge es i diffuse la ge B cell lymphoma,” said Me edith Nix, Ph.D. ca didate at Sta fo d U ive sity a d study co-autho . “It was g atifyi g to see that Deep O igi ’s computatio al a alysis offe ed a atio ale fo why TCIP3 was the most effective KAT-TCIP at activati g these ep essed t a sc iptio al p og ams.”
The computatio al plus i vivo data p ovide evide ce of Deep O igi ’s platfo ms’ biological p edictivity i livi g systems. Lea mo e about the Deep O igi platfo m at https://www.deepo igi .com/platfo m.
To ead the full esea ch publicatio i Cell, visit: https://www.cell.com/cell/abst act/S0092-8674(26)00757-9.
Deep O igi is buildi g computatio al discove y systems that model life, to close the gap betwee p ecli ical p edictio s a d cli ical outcomes. Co-fou ded by Michael A to ov, co-fou de of Oculus, a d Ga egi Papoia , Ph.D., Mo oe Ma ti P ofesso of biochemist y at the U ive sity of Ma yla d, Deep O igi u s d ug discove y p og ams — its ow a d its pa t e s’ — th ough hyb id AI-mecha istic models that spa biological scales f om qua tum to cellula to huma body scale. The compa y e gages th ough discove y pa t e ships a d SaaS platfo m access. Deep O igi is backed by mo e tha $50 millio i capital a d mo e tha $32 millio i co t acts, i cludi g a $30 millio ARPA-H CATALYST awa d to build i silico models that ca eplace a imal testi g i p ecli ical d ug developme t. Fo mo e i fo matio , visit https://deepo igi .com/.