Biology 2026

Lentiviral and retroviral gene therapies permanently insert DNA semi-randomly into the genome. If they land near a cancer gene, they can turn it on — this caused leukemia in 6 of 20 patients in early X-SCID trials. CAR-T therapies face the same risk. FDA requires 15 years of post-treatment monitoring.

Nearly all approved gene therapies treat single-gene diseases. But the diseases that kill the most people — heart disease, diabetes, Alzheimer's, most cancers — involve dozens to hundreds of genes. We don't know how to safely edit multiple genes simultaneously, and multiplexed editing exponentially increases off-target risks.

Base editors (~5.2 kb) and prime editors (~6.3 kb) exceed AAV's 4.7 kb packaging limit. Split-intein dual-AAV approaches work but lose 40-70% of activity. This means the safest, most precise editing tools can't use the most established delivery vehicle.

Precise gene correction via homology-directed repair (HDR) only works in dividing cells. But the cells you most want to fix — brain neurons, heart muscle, skeletal muscle — are non-dividing, so they default to error-prone repair that creates random mutations instead of precise corrections. Prime editing helps but achieves only 1-10% efficiency in vivo.

Both AAV and LNP vectors naturally end up mostly in the liver after IV injection. This is great for liver diseases but means lungs, kidneys, pancreas, and immune cells receive very little therapy. Cystic fibrosis lung gene therapy has failed for 30 years because of this delivery problem.

CRISPR sometimes cuts at wrong locations in the genome that look similar to the target. Beyond point mutations, these unintended cuts can cause large deletions (megabases), inversions, and translocations — potentially activating cancer genes. This is the biggest regulatory bottleneck for expanding CRISPR therapies.

30-70% of the population has pre-existing antibodies against AAV from natural childhood exposure, making them ineligible for gene therapy. Worse, patients who receive one dose develop strong immunity preventing a second dose — even if the first wears off years later.

The most commonly used gene therapy delivery vehicle (AAV) can only carry ~4,700 base pairs of DNA. But many serious diseases involve much larger genes: dystrophin for Duchenne muscular dystrophy is 14 kb, CFTR for cystic fibrosis needs ~6 kb with regulatory elements, ABCA4 for Stargardt disease is 6.8 kb.

After cancer surgery or chemotherapy, tiny amounts of remaining cancer (minimal residual disease or MRD) often recur months to years later. ctDNA blood tests can detect MRD, but sensitivity varies by cancer type, and it's unclear whether MRD-guided treatment decisions improve outcomes compared to standard protocols.

Tuberculosis kills 1.3M people/year (2nd deadliest infectious disease after COVID). In low-resource settings where 95% of TB deaths occur, the primary diagnostic (sputum smear microscopy) misses 40-60% of cases, especially in HIV co-infected and pediatric patients.

Ovarian cancer is the deadliest gynecological cancer because there is no effective screening test. CA-125 blood test and transvaginal ultrasound have been tested extensively but failed to reduce mortality in large RCTs. 70%+ of cases are diagnosed at Stage III/IV where 5-year survival is only 30%.

Sepsis kills ~270,000 Americans annually and is the most expensive hospital condition. Current biomarkers (CRP, procalcitonin, IL-6) lack specificity and take hours to return results. There is no rapid point-of-care test that reliably detects sepsis before organ failure begins.

The first FDA-cleared blood test for Alzheimer's (Lumipulse G p-tau217/Aβ1-42 ratio, May 2025) achieves 91.7% concordance with amyloid PET scans. But it's only cleared for symptomatic patients 55+, and real-world performance in diverse populations, primary care settings, and early pre-symptomatic stages remains unclear.

Pancreatic cancer is diagnosed late in >80% of cases because there is no screening test for the general population. 5-year survival is below 12% overall, but screening programs report 24-73% 5-year survival when caught early. A new 4-biomarker blood test (ANPEP, PIGR, CA19-9, THBS2) achieved 91.9% accuracy in a 2026 study.

GRAIL's Galleri test can detect 50+ cancer types from a blood draw, but sensitivity varies dramatically by cancer stage and type — high for late-stage cancers but potentially missing many early-stage ones. GRAIL submitted FDA PMA in January 2026 with data from 25,490 PATHFINDER 2 participants and 140,000+ NHS-Galleri trial participants. Over 475,000 commercial tests sold.

When a patient has a bacterial infection, doctors must prescribe antibiotics immediately without knowing which antibiotic will work. Traditional culture-based susceptibility testing takes 2-3 days. This forces broad-spectrum empiric prescribing that drives antibiotic resistance.

Gene therapies are the most expensive drugs ever approved — Hemgenix costs $3.5M, Elevidys $3.2M, Zolgensma $2.25M per patient. Manufacturing relies on artisanal processes (transient transfection of adherent cells) that cannot scale. Global AAV production can serve only ~5,000-10,000 patients per year.

Despite $50B+ invested in AI drug discovery companies, not a single AI-designed drug has been approved by the FDA. Several AI-discovered candidates are in Phase II, but none have completed Phase III. The validation gap between computational predictions and clinical outcomes remains wide.

Biologic drugs (antibodies, peptides, proteins) represent ~40% of pharma revenue but almost all require injection because stomach acid and enzymes destroy them. Oral semaglutide (Ozempic) is a rare exception but requires a special absorption enhancer and achieves only ~1% bioavailability.

More than 90% of drugs that work in animals fail in human trials. Mouse models of cancer, Alzheimer's, and inflammation are notoriously poor predictors. This is the fundamental reason clinical trials fail at such high rates and cost billions.

Of approximately 10,000 known rare diseases, only 5% have any treatment. Healthcare expenses for Americans with rare diseases are 3-5x greater than for those without. The total economic burden approaches $1 trillion annually. Small patient populations make traditional drug development ROI calculations impossible.

Unlike other drugs, new antibiotics are deliberately used sparingly to prevent resistance, which destroys the economic incentive to develop them. Multiple antibiotic startups have gone bankrupt despite FDA approval. The PASTEUR Act proposes $6B in subscription-style payments to fix this, but hasn't passed yet.

The blood-brain barrier prevents most drugs from reaching brain cells. Treating neurological diseases requires invasive brain surgery for direct injection, which is risky and cannot treat the whole brain uniformly. Focused ultrasound with microbubbles is emerging but remains early-stage.

More than 30% of drugs fail in Phase II and 58%+ fail in Phase III. CNS drugs fail at ~85%. The average Phase III trial costs $36.58M (2024), 30% higher than 2018. The overall success rate from Phase I to approval is approximately 8-11%, meaning 9 out of 10 compounds fail.

Cells maintain protein quality through chaperones, the proteasome, and autophagy. This system collapses with age, allowing misfolded proteins to accumulate — causing Alzheimer's (amyloid/tau), Parkinson's (alpha-synuclein), and cataracts. ISRIB dramatically improved memory in aged mice but has not yet translated to humans.

The thymus — the organ that trains new immune cells — shrinks dramatically after puberty, losing 90%+ of functional tissue by age 50. Old people cannot mount effective immune responses to new infections or vaccines. Greg Fahy's TRIIM trial (N=9) showed partial thymic regeneration, but no properly powered trial has confirmed this.

Mitochondria accumulate DNA mutations at 10-17x the rate of nuclear DNA due to poor repair mechanisms. Damaged mitochondria can outcompete healthy ones within a cell through 'clonal expansion.' There is no approved therapy to selectively remove or repair mutant mtDNA, leaving cells increasingly energy-starved with age.

Rapamycin is the most consistently life-extending drug ever found in lab animals — working in yeast, worms, flies, and mice (10-25% lifespan extension). But it was designed as an immunosuppressant. The Dog Aging Project TRIAD trial ($7M NIH grant, 580 dogs) won't conclude until 2029, and ARPA-H just funded a $30.8M human trial of a next-gen rapamycin analog.

NAD+ is a critical molecule that declines 40-60% between ages 30 and 70. Supplements like NMN and NR reliably double circulating NAD+ levels in humans within 14 days, but clinical trials have not consistently shown the dramatic health improvements seen in mice.

Scientists can reset old cells to a younger state using Yamanaka factors, but pushing too far turns cells cancerous or makes them forget their identity. The first human trial of epigenetic reprogramming therapy was FDA-cleared in 2025 (Life Biosciences), but nobody has established the precise 'dose' that rejuvenates without causing tumors.

Senescent 'zombie cells' accumulate with age and poison surrounding tissue. Drugs that kill them (senolytics) show dramatic benefits in mice, but current drugs like dasatinib+quercetin are repurposed cancer drugs that also damage healthy cells — particularly platelets and immune cells — leading to inconsistent clinical results.

We lack a reliable, FDA-accepted way to measure biological age. Without this, testing anti-aging interventions requires decades-long trials waiting for disease or death. Epigenetic clocks like GrimAge and DunedinPACE show promise but disagree with each other and have not been validated as regulatory endpoints.