Base editors fuse a CRISPR nickase to a single-stranded DNA deaminase, enabling precise single-nucleotide changes without double-strand breaks, and in principle could correct ~62% of known pathogenic SNVs. However, highly evolved adenine base editors such as ABE8e and ABE8.20 trade precision for power: their broadened editing windows (positions 3–9 or 3–11) drive extensive bystander editing and elevated guide-dependent and guide-independent off-target activity in DNA and RNA.

The Nature study asks whether we can keep ABE8-level on-target activity while restoring tight control over where edits occur. Through systematic “reversion analysis” of ABE7.10’s 14 mutations, the authors build minimally evolved ME-ABEs that match ABE8e/8.20 activity at central positions yet retain narrow ABE7.10-like windows and >10-fold lower unguided off-target DNA editing.

These editors better satisfy the field’s goal of high on-target activity, low off-target activity and minimal window size, positioning ME-ABEs as strong candidates for disease-gene correction and as a template for refining other base editors.

Merck’s decision to scrap its planned GBP £1B London R&D hub, cutting 125 science jobs, has become a bellwether for pharma’s fraying confidence in the UK. Sir John Bell, a key architect of the COVID-19 vaccine rollout, says multiple big‑pharma CEOs now see Britain as uninvestable, citing chronic underinvestment and aggressive revenue clawbacks on branded drugs.

The base problem is that UK spends just 9% of healthcare budgets on medicines versus 14-20% across the OECD, while rebate rates far exceed France and Germany. For companies weighing global pipelines, this erodes returns on innovative therapies. Bell argues that without anchor pharma, the UK’s world‑class academia and biotech cannot scale, urging creative fixes – from targeted funds to fairer pricing – to stem an emerging exodus.

The US FDA has released draft guidance that could accelerate a shift away from animal testing in drug development. The document outlines how drug makers should validate “new approach methodologies” (NAMs) such as advanced in vitro systems, organoids, organs-on-chips, in silico models and lower organisms, so these tools can stand in for traditional animal toxicology studies. The move is important because validated human-relevant NAMs can better predict clinical responses, de-risk early trials and cut timelines, cost and ethical concerns tied to animal use.

The guidance centres on four validation pillars - context of use, human biological relevance, technical characterisation and fit-for-purpose performance - signalling regulators’ readiness to accept NAMs as primary evidence, and paving the way for more human-centric, data-rich R&D pipelines.

Roche is ramping up its AI ambitions with a hybrid‑cloud “AI factory” built on 3,500 NVIDIA GPUs, making it the pharma sector’s largest known deployment of this kind of infrastructure. The concept is to treat compute as a production line: continuously training and deploying generative models across R&D, manufacturing, diagnostics and digital health.

Strategically, this matters because AI‑native platforms like Nvidia’s BioNeMo can shorten hit discovery, optimise candidates and simulate trials, boosting R&D productivity and differentiation against peers like Lilly. Early efforts focus on Roche unit Genentech’s Lab‑in‑the‑Loop to bring generative AI to drug discovery and digital pathology. Longer term, Roche is signalling that AI‑accelerated pipelines and data‑centric operations will be core to its competitive edge in therapeutics and diagnostics.