Why some people are more resistant to developing blood cancer even when they carry cancer-risk mutations

Blood cancer is an umbrella term for a variety of diseases that affect the blood, bone marrow and lymphatic system. Like most cancers, the cause is usually mutations in the DNA, which are genetic errors that accumulate as we age. However, some people are more resistant to developing blood cancer even when they carry cancer-risk mutations.

In a new study published in the journal Science, researchers discovered why. They identified a rare genetic variant that reduces the risk of several blood cancers, including leukemia, by slowing down a process called clonal hematopoiesis (CH).

CH occurs when a hematopoietic stem cell, which can develop into any type of blood cell, mutates and grows into a large population of identical mutated cells.

In their study, the team performed a large-scale analysis (known as a GWAS meta-analysis) of more than 640,000 individuals. By comparing 43,000 people with CH mutations to 600,000 without them, they pinpointed the genetic variant that protects against CH.

The researchers identified a noncoding regulatory variant, rs17834140-T, on chromosome 17q22.

The experiments revealed that rs17834140-T reduces the amount of a protein called MSI2, which acts as a growth booster in stem cells. In cancer, MSI2 causes mutated cells to multiply rapidly and take over the bone marrow. However, the protective variant results in lower levels of the protein, forcing these mutated cells to grow much more slowly and reducing their chances of progressing to leukemia. According to the study authors, people with the variant have up to a 30% lower risk of developing CH.

The research could lead to new ways to prevent cancer even before it starts. Now that we know that lowering MSI2 is protective, scientists may be able to develop drugs or other therapeutic approaches that mimic or enhance this natural protection.

Gaurav Agarwal et al, Inherited resilience to clonal hematopoiesis by modifying stem cell RNA regulation, Science (2026). DOI: 10.1126/science.adx4174

Francisco Caiado et al, Genetic resistance to leukemia, Science (2026). DOI: 10.1126/science.aed5244

Tire rubber decays into a potentially dangerous chemical cocktail, research shows

Crumb rubber from recycled tires, commonly used in artificial turf, generates hundreds of transformation chemicals as it decays under sunlight and environmental conditions. Some of these chemicals, such as 6PPD-quinone, are highly toxic to aquatic life, while others are known human health hazards. The long-term effects of most transformation products remain unknown.

The small, spongy black beads used as fill material in most artificial turf fields are called crumb rubber, which has long been touted as a major win for recycling. However, conflicting studies have alternately identified crumb rubber as either safe for people to play atop or dangerous to human health.

New research out of Northeastern University investigated the decay cycle of crumb rubber, which is fashioned out of old tires. By simulating the conditions in which the rubber decays, like strong sunlight, they discovered that crumb rubber is highly reactive, generating hundreds of previously untracked chemicals as it decays, some of which are hazardous to humans.

The work is published in the journal Environmental Science & Technology.

Madison H. McMinn et al, From the Road to the Field: Decoding Chemical Transformation in Aging Tire and Artificial Turf Crumb Rubber, Environmental Science & Technology (2025). DOI: 10.1021/acs.est.5c08260

AI shows bias against some Indian castes

Popular large language models (LLMs) often reproduce harmful stereotypes about Indian castes — hereditary groups traditionally associated with specific occupations and social status. Researchers used a custom-designed tool to detect ‘caste bias’ in LLMs and found that every model they tested exhibited some bias. GPT-4o and GPT-3.5, created by OpenAI, had some of the highest bias scores. Information on minority groups might be less likely to appear in prestigious journals or other outlets, and might be written in local languages, which could result in it being filtered out of AI training data who studies cultural biases in LLMs.

[2508.03712] How Deep Is Representational Bias in LLMs? The Cases o...

[2510.02742] IndiCASA: A Dataset and Bias Evaluation Framework in L...

[2505.14971] DECASTE: Unveiling Caste Stereotypes in Large Language...

AIs are biased toward some Indian castes — how can researchers fix ...