GARBH-INi builds AI tools to predict preterm births

India's largest pregnancy cohort of about 12,000 women anchors a DBT-led maternal-health research programme building India-specific tools to predict and date preterm births.

What happened

• The Department of Biotechnology (DBT) held a dissemination programme on the learnings and outcomes of GARBH-INi — the Interdisciplinary Group for Advanced Research on Birth Outcomes — at the India Habitat Centre, New Delhi.
• The event was addressed by Union Minister of State (Independent Charge) for Science & Technology and Earth Sciences, Dr. Jitendra Singh, with the Secretary of Biotechnology, a NITI Aayog Member, and the Executive Director of THSTI in attendance.
• The Minister said GARBH-INi has enrolled around 12,000 pregnant women, creating one of South Asia's largest pregnancy cohorts, to develop indigenous, AI-driven solutions for preterm birth — a leading cause of neonatal death and of long-term adult morbidity.
• A compendium documenting the programme's key learnings and outcomes was released.
• Technology transfers and partnerships were formalised: a microbiome-based biotherapeutic technology was licensed to a private firm, and Letters of Intent were signed with two health-technology companies for AI-enabled ultrasound reporting and risk-stratification under the GARBH-INi-AnandiMaa initiative.
• The Minister framed maternal and child health as central to India's growth toward its 2047 development goals, noting that children born today will shape the country's future productivity.

Background & context

What "preterm birth" means. A birth is preterm when it occurs before 37 completed weeks of pregnancy. Preterm babies face higher risks of death in the newborn period and of lasting health and developmental problems into adulthood. India carries a significant share of the global preterm-birth burden, which is why the Minister stressed that the solutions must be built for Indian conditions rather than imported wholesale — gestational-age and growth standards derived from Western populations do not map cleanly onto Indian mothers and babies.

Where GARBH-INi sits. GARBH-INi is a programme of the Department of Biotechnology (DBT), which is one of the departments under the Ministry of Science & Technology (the other two being the Department of Science & Technology, DST, and the Department of Scientific & Industrial Research, DSIR). It is implemented through the Translational Health Science and Technology Institute (THSTI), a DBT autonomous institute at Faridabad in the National Capital Region. The acronym is literal: GARBH-INi = Interdisciplinary Group for Advanced Research on Birth Outcomes, the word garbh evoking the womb. The model is a longitudinal cohort study — the same set of pregnant women is followed forward through pregnancy and after delivery so that researchers can link what was measured early on to the birth outcome that actually occurred.

Why a cohort and a biorepository. Predicting a rare, multi-factor outcome such as preterm birth needs scale and depth together: many women followed closely, with biological samples and imaging banked at fixed points. GARBH-INi pairs clinical epidemiology (who, when, under what conditions) with multi-omics biomarkers (genomic, microbiome and other molecular signals) and artificial intelligence to turn that data into personalised risk predictions. This combination — population-scale clinical data plus a banked biospecimen repository plus AI models — is what distinguishes it from a one-off clinical trial.

The three building blocks, explained. First, clinical epidemiology records each pregnancy in standardised detail — visits, measurements and conditions — so outcomes can be traced back to early exposures. Second, multi-omics means reading several layers of biology at once: the genome (inherited DNA variants), and the microbiome (the community of microbes a woman carries), among others; GARBH-INi's finding of microbiome-based predictors says that the maternal microbial profile itself carries signal about preterm risk. Third, artificial intelligence is the layer that learns patterns across thousands of pregnancies and millions of images to estimate risk for a new individual. The ultrasound images matter especially for dating: how far along a pregnancy is drives almost every clinical judgement that follows, and a model trained on Indian scans can date more accurately than formulas built abroad.

For Prelims

• Full form: GARBH-INi = Interdisciplinary Group for Advanced Research on Birth Outcomes.
• Parent & implementer: a Department of Biotechnology (DBT) initiative, run through the Translational Health Science and Technology Institute (THSTI), a DBT autonomous institute at Faridabad.
• Ministry chain: Ministry of Science & Technology → Department of Biotechnology → THSTI.
• The cohort: around 12,000 pregnant women — one of South Asia's largest pregnancy cohorts.
• The repository: over 1.6 million well-characterised biospecimens and more than 1 million ultrasound images.
• Stated method: integrates clinical epidemiology, multi-omics biomarkers, and artificial intelligence for personalised predictions.
• Named outputs: AI-based pregnancy-dating models tailored for Indian populations; microbiome-based predictors of preterm birth; rapid diagnostic tools; and genetic markers for early risk assessment.
• Platforms built: a national biorepository and the GARBH-INi-DRISHTI data-sharing platform for the research community.
• AnandiMaa track: the GARBH-INi-AnandiMaa initiative covers AI-enabled ultrasound reporting and risk-stratification, advanced via technology transfer and Letters of Intent with private health-tech firms.
• Problem addressed: preterm birth (delivery before 37 weeks), a leading cause of neonatal mortality and of adult-life morbidity, with India bearing a large share of the global burden.
• Document released: a compendium of the programme's learnings and outcomes.

What it is NOT. GARBH-INi is not a welfare scheme that transfers money or services to mothers, and it is not a hospital network — it is a research programme and biorepository. It is run by DBT/THSTI, not by the Ministry of Health and Family Welfare, and is distinct from health-ministry maternal programmes such as Janani Suraksha Yojana, Pradhan Mantri Surakshit Matritva Abhiyan or POSHAN Abhiyaan, which deliver services and nutrition rather than build AI prediction models. It is also distinct from DST (the sibling department that funds the physical sciences) — GARBH-INi belongs to DBT. "DRISHTI" here is GARBH-INi's data-sharing platform and should not be confused with any unrelated programme of the same name.

The DBT family it sits within. For "match the institution" and "how many of these are under DBT" questions, GARBH-INi's home institute THSTI is one of DBT's autonomous research institutes; the wider DBT ecosystem includes bodies such as the National Institute of Immunology (NII), the National Centre for Cell Science (NCCS), inStem, the Regional Centre for Biotechnology (RCB), and the Biotechnology Industry Research Assistance Council (BIRAC). GARBH-INi's biorepository-plus-AI approach is the maternal-health counterpart to other national biobanking and genomics efforts; the broad point for the exam is that it is part of India's growing bioeconomy and translational-health push.

Why it matters

Preterm birth is both a humanitarian and a development problem. As a leading cause of newborn death, it weighs directly on India's infant- and neonatal-mortality indicators; and because survivors carry elevated risk of long-term illness, the burden extends across a lifetime into adult productivity. The Minister's framing — that children born today will define the country's strength in 2047 — ties maternal and child health to the demographic-dividend argument: a healthy birth cohort is human capital.

The deeper significance is methodological. Clinical tools used in Indian obstetric practice — including the ultrasound formulas that estimate how far along a pregnancy is — were largely calibrated on non-Indian populations, so they can mis-date pregnancies and mis-classify risk. By building pregnancy-dating models tailored to Indian populations from a 12,000-woman cohort, GARBH-INi addresses a quiet accuracy gap that affects every downstream decision in antenatal care. The microbiome and genetic-marker work aims to move prediction earlier, so that high-risk pregnancies can be flagged and managed before complications set in rather than after.

Finally, the programme demonstrates a model that India increasingly relies on: a publicly funded cohort and biorepository generating data and prototypes, then handing validated technologies to industry through licensing and Letters of Intent for scale-up — here, microbiome-based biotherapeutics and AI-enabled ultrasound reporting. The GARBH-INi-DRISHTI data-sharing platform extends that further by opening characterised data to the wider research community, which is how a single cohort can seed many downstream studies and publications.

How it compares to a clinical trial. A randomised clinical trial tests one intervention on a smaller group to ask "does this treatment work?" GARBH-INi is the opposite kind of instrument: an observational cohort that follows a large group as they are, asking "what predicts this outcome, and how early can we see it?" That is why its products are predictors, risk markers and dating models rather than a single drug or device. The trade-off is that a cohort cannot by itself prove a treatment causes an effect; its strength is discovery and prediction at population scale, which then feeds the more focused trials and the diagnostic tools that companies can commercialise.

The translation layer. The specific deals announced make the pipeline concrete. The microbiome-based biotherapeutic technology — a product line that flows directly from the microbiome predictor finding — was transferred to a private probioceuticals company, moving a research insight toward a manufacturable product. The two Letters of Intent, for AI-enabled ultrasound reporting and for risk-stratification platforms under the AnandiMaa track, point the imaging and prediction work toward deployable clinical software. Together with the felicitation of participating families, the programme also signals something quieter: long-run cohorts depend on the sustained, voluntary participation of ordinary citizens, whose biospecimens and repeat scans are the raw material the AI is built on.

For Mains