Research Blends DNA, AI To Fight Cancer

“Cancer cannot be explained by mutations alone. Today, cancer is understood as a multifactorial disease, shaped by genetics, gene regulation, environment and time,” said Nita Parekh, Professor of Bioinformatics at IIIT‑H.

A century ago, scientists believed cancer began with a single mistake inside a cell. The somatic mutation theory of 1914 proposed that abnormalities in DNA could trigger uncontrolled growth. That explanation widened as oncogenes and tumour suppressor genes were discovered, and as evidence grew that tissue environment, viruses and cellular stress influence tumour development.

Prof. Parekh’s work examines variations in DNA and how they drive tumorigenesis. Some changes involve a single letter in the genetic code, while others include missing segments, duplications, inversions or even gene fusions across chromosomes. These variations can accumulate silently or alter cell behaviour. “By analysing cancer genomes in detail, we can identify which mutations matter, which pathways they disrupt, and how different cancers — or even subtypes of the same cancer —behave very differently,” she explained.

One focus of her research is diffuse large B‑cell lymphoma, an aggressive blood cancer that accounts for about 37 per cent of non‑Hodgkin’s lymphoma. The disease has two main subtypes and outcomes vary widely. “When we looked at subtype‑specific genetic variations, we could clearly see why some patients respond well to treatment while others do not,” she said. Her team identified subtype‑specific mutations, disrupted pathways and biomarkers that guide treatment choices based on a patient’s mutational profile.

Genes alone do not tell the full story, Prof. Parekh added. Epigenetic switches such as DNA methylation can silence tumour‑suppressing genes or activate cancer‑promoting ones without altering the DNA sequence. Her group studies methylation across promoters, enhancers, gene bodies and non‑coding RNAs. “Epigenetic regulation is the ongoing part of my work,” she said.

Her research also extends to breast cancer, where DNA methylation data, RNA expression profiles and machine learning have been used to identify molecular signatures that distinguish subtypes and predict risk and survival. Findings point to liquid biopsies that detect cancer signals in blood before symptoms appear.

Prof. Parekh is also working on mammography data. Her team has built curated datasets to train AI models that detect abnormalities early, classify tumours and generate preliminary reports. This work supports radiologists and reduces delays, particularly in resource‑constrained settings where early diagnosis remains a challenge.