Al­though Sum­mit Ther­a­peu­tics didn’t present any da­ta at AS­CO's an­nu­al meet­ing, it still dom­i­nat­ed the con­ver­sa­tion at Mc­Cormick Place in Chica­go.

The com­pa­ny put out a press re­lease the same Fri­day morn­ing that the con­fer­ence start­ed in a games­man-like move, out­lin­ing new da­ta for its PD1xVEGF bis­pe­cif­ic called ivonescimab and out­shin­ing many of the late-break­ing and ple­nary pre­sen­ta­tions. An­a­lysts said Sum­mit went on to host sev­er­al meet­ings in Chica­go, as in­vestors grap­pled with a very im­por­tant ques­tion: How ma­ture were the over­all sur­vival da­ta in this spe­cif­ic lung can­cer pop­u­la­tion?

Ac­cord­ing to an­a­lyst notes over the last few days, Sum­mit's po­si­tion is that the sta­tis­ti­cal­ly not-sig­nif­i­cant OS fig­ure it re­port­ed last week is the fi­nal analy­sis. That could pose a chal­lenge to ap­proval as the FDA will re­quire a sta­tis­ti­cal­ly sig­nif­i­cant re­sult, per Sum­mit’s re­lease.

Re­verse merg­ers have once again lost their lus­ter.

An al­ter­na­tive to go­ing pub­lic, re­verse merg­ers once car­ried a neg­a­tive stig­ma. But the fi­nanc­ing ve­hi­cle gained new­found at­ten­tion in re­cent years as the biotech IPO mar­ket soured and pri­vate com­pa­nies lined up to take the place of down-and-out pub­licly-trad­ed drug de­vel­op­ers.

At least 16 such deals were an­nounced in 2024, ac­cord­ing to a tal­ly from William Blair bankers. One com­pa­ny — Vin­cerx Phar­ma — tried two dif­fer­ent re­verse merg­ers be­fore giv­ing up. It de­cid­ed to wind down in April.

Few­er than five re­verse merg­ers have been dis­closed so far in 2025, in­clud­ing the planned com­bi­na­tions of Salarius Phar­ma­ceu­ti­cals with De­coy Ther­a­peu­tics and Eye­n­ovia with Be­taliq.

A trio of young MIT sci­en­tists un­veiled a new AI bio mod­el on Fri­day that pre­dicts bind­ing affini­ties, a ma­jor chal­lenge to dis­cov­er­ing drugs that has pre­vi­ous­ly evad­ed progress with oth­er AI meth­ods.

Boltz-2 takes a step be­yond the May 2024 re­lease of Al­phaFold 3, the lat­est gen­er­a­tion of Google Deep­Mind's mod­el that pre­dicts bio­mol­e­c­u­lar struc­tures. Boltz-2 pre­dicts struc­tures like Al­phaFold, but it al­so pre­dicts bind­ing affin­i­ty, or how strong­ly a mol­e­cule binds to a tar­get pro­tein.

“Bind­ing affin­i­ty is core to de­vel­op­ing a ther­a­peu­tic, start to fin­ish,” Re­cur­sion R&D head Na­jat Khan said. (Re­cur­sion helped train Boltz-2 with its com­pute pow­er and bench­mark the mod­el.) “It’s been the fun­da­men­tal is­sue that a lot of us have been try­ing to grap­ple through.”