Immune cells called macrophages make up a large chunk of a glioblastoma tumor — and they help it grow, hide from treatment, and resist therapy. The problem is that standard lab methods for studying these cells don’t reflect their real complexity. This dataset introduces a new co-culture model using actual patient tumor cells to train macrophages more realistically. The resulting cells match immune subtypes seen in real patient tumors far better than existing methods, and ones grown alongside radiation-treated tumor cells even mirrored patterns from recurrent disease. A key takeaway: gene activity alone doesn’t always predict how cells behave, so functional testing still matters.
