4baseCare

Two businesses sit under one name, and the AI leads only one of them. Most of the revenue comes from a network of hospital linked sequencing laboratories running the TARGT panels, which is laboratory diagnostics rather than AI. OncoTwin is the learned layer: it matches a patient to clinically and genomically similar patients using tuned similarity weights and graph methods, then reports outcome rates across the matched cohort.

The whole exome tests add a therapy response index produced by Cellworks, a US biosimulation company. Remove the models and the tests still sell; remove them from OncoTwin and nothing is left. The AI is the engine of one core module.

Decision support, with the oversight structure described in some detail. The OncoTwin page states that the output is decision support rather than a clinical directive, that its outcome figures are observational rather than causal, that the rationale behind each match is shown, and that small cohorts are flagged. Each report carries annotations for sparse data, confounding and completeness, and the company states that every genomic report passes three reviews before release: the bioinformatic pipeline, a molecular geneticist and a clinical oncologist. Cases can also be taken to the company's virtual molecular tumor board.

Two parts are missing: a threshold at which a match is withheld rather than flagged, and what happens after a recommendation proves wrong. The marketing pulls against the guardrails in places, describing digital twins that predict response and simulate therapy options, and the Cellworks report on the whole exome tests ranks the top 25 therapies by predicted response.

The approach and the supplier are both named. OncoTwin is described step by step: clinical, biomarker and genomic data are normalized, candidate cohorts are constrained by stage and driver to avoid implausible comparisons, matches are scored with tuned weights, subcohort logic and co-alteration penalties, and each match is explained with its drivers and a confidence band. Graph and network methods are named for the similarity layer. For the whole exome tests the model is Cellworks' Computational Omics Biology Model, which simulates pathway response from sequencing and transcriptome data and ranks therapies by a Therapy Response Index in a Cellworks Ventura report.

Neither layer publishes a version, a description of the data the matching weights were tuned on, or an update practice, and the OncoTwin page calls its evidence engine continuously updated, so a buyer cannot tell when the system changed.

One model supplier is named. For the whole exome tests, TARGT Absolute Plus and Absolute SoLiQ, the brochure states that the Cellworks biosimulation platform takes the patient's sequencing and transcriptome data and returns a Cellworks branded report ranking therapies; Cellworks is a US company, named on the product page and in the brochure. OncoTwin and the laboratory pipelines are described as in house.

The chain is not enumerated beyond that. The privacy policy refers to a cloud platform on a secured server whose provider is not named, and to third party service providers under confidentiality agreements without listing them. No terms are published for what Cellworks may do with exome data once it arrives, or for where that step is processed.

Published research on the assays, nothing yet on the AI. The whole exome DNA and RNA workflow behind the Absolute tests was validated in a peer reviewed study (Frontiers in Oncology, 2023) of 166 patients across 17 cancer types, reporting a limit of detection of 5 percent for single nucleotide variants and 10 percent for insertions and deletions, 100 percent sensitivity, 97.5 percent specificity and more than 98 percent concordance with orthogonal methods. The authors are company employees with one Tata Memorial Centre coauthor, so the study is published but not independent. Further company papers describe tumor mutation burden, gene fusions and other alterations in Indian cohorts.

No study tests whether OncoTwin's matched cohort outcomes or the Cellworks therapy ranking change treatment choices or results for these patients. Cellworks publishes its own studies of its therapy response index, but none were run on this product's reports. Patient volumes and selection into Memorial Sloan Kettering's innovation hub cohort for 2025 are scale and recognition rather than performance.

A privacy policy reaches the service, and what it permits runs against the patient. It states that personal and sensitive data, including medical records and genetic samples, may be used for commercial purposes and, in aggregated and anonymized form, for research and business intelligence, and that such research, statistical or intelligence data may be sold or transferred to third parties and affiliates. Personal data is kept until the person emails an opt out address after the relationship ends, subject to legal holds.

The company's stated strategy is a large longitudinal clinico genomic database, the Global Cancer Diversity Atlas, feeding an evidence engine updated continuously from real world data, so what counts as anonymized and who receives it carry real weight; neither is defined. Nothing states whether patient data trains OncoTwin or what reaches the Cellworks model beyond producing the report. Nothing is published on safety engineering either, meaning guardrails, hallucination mitigation or how a safety event is handled.

Outside the United States, with an equivalent regime documented. The laboratories serve patients and hospitals in India, the United Arab Emirates, Nepal and the Philippines, and no business associate position is taken because HIPAA does not reach that processing. The privacy policy is issued under India's Information Technology Act and its 2011 rules on sensitive personal data, lists access, correction, erasure, restriction and portability rights with reference to GDPR, states a retention rule and a breach notice commitment, and covers the medical records and genetic samples handled through the service.

It does not cite India's Digital Personal Data Protection Act of 2023 or the data laws of the other three countries, and a United States buyer would find no agreement to sign.

The security controls are the company's own claims, with nothing independent behind them. The privacy policy states SSL encryption in transit for certain information, storage on a secured server whose unnamed provider says it follows best security practice, and procedures in line with good industry practice. No SOC 2 report, ISO/IEC 27001 certificate, HITRUST certification, penetration testing cadence or trust portal is published.

The accreditation the company does hold, College of American Pathologists accreditation for its Dubai laboratory announced in March 2025, audits analytical quality and laboratory operation rather than information security, so it is not a security certification.

No United States device claim is made, and the products are sold outside the United States. The tests are laboratory developed assays run in the company's own accredited laboratories, including College of American Pathologists accreditation for the Dubai laboratory, and OncoTwin is scoped on its own page as decision support rather than a clinical directive, with outcome figures described as observational rather than causal. That positioning fits the absence of a clearance.

No CDSCO or other national device registration is stated for OncoTwin or for the Cellworks therapy ranking, a gap that matters wherever decision support software is regulated as a device.

Published results within the population the product is built for, short of a comparison of model performance across groups. The company's thesis is that genomic references and trials over-represent white patients, and it acts on it: the Indiegene panel draws on gene discoveries from South Asian tumors, and a peer reviewed study of 1,233 Indian patients across 30 cancer types (JCO Global Oncology, 2023) found tumor mutation burden thresholds that differed from The Cancer Genome Atlas in sarcoma, ovarian, head and neck and breast cancers but not in lung or colorectal cancer. The product now reports a cancer and population adjusted percentile. That is evaluation within the group where a Western threshold is most likely to misclassify.

Nothing compares how OncoTwin's matches or the Cellworks therapy ranking perform across populations, cancer types or levels of data completeness.

Mechanisms exist to challenge an output, and nothing stands behind it. Each OncoTwin match shows its drivers, the cohort size behind each outcome rate and a confidence band, the report annotates sparse data and confounding, and genomic reports pass pipeline, molecular geneticist and oncologist review before release, so a clinician can test a recommendation before acting on it.

The terms of use run the other way: they state the platform may not be error free, disclaim responsibility for third party providers, and have the user indemnify the company. No error rate for the matching or the therapy ranking is published, and no route exists for a patient or clinician to hold the company to a recommendation that proves wrong.

No integration evidence. No hospital record system, interface standard or integration partner is named anywhere on the site. OncoTwin's tumor board workspace consolidates genomic findings, clinical history and prior treatments in its own format, and the OncoBuddy patient app stores records that patients upload themselves. Sequencing capacity is placed inside partner hospitals, which embeds the company in the care pathway physically but says nothing about how results reach the chart.

Options and residency are stated, with the processing path left open. Testing runs in the company's own laboratories in India, the United Arab Emirates, Nepal and the Philippines, several placed inside partner hospitals so that samples do not travel to a central laboratory, and the privacy policy states the platform is located in India.

The cloud platform that holds medical records and runs OncoTwin sits with a provider that is not named, tenant isolation is not described, and the Cellworks step in the whole exome tests is run by a US company with no stated location of processing.

No price is published, and the posture is discoverable. Tests are ordered per patient by the treating oncologist, the portfolio page sets out which test fits which case, and brochures and sample reports can be requested from it. OncoTwin is offered to oncologists, hospitals and pharmaceutical sponsors with no stated charge, and whether it is sold apart from testing is not stated. A free testing program for eligible lung cancer patients in India runs with a partner foundation, and a hospital program targets affordability; neither states a price.

Coverage is named, with validation behind part of it. The product is oncology only: solid tumor tissue and liquid biopsy profiling, combined tissue and blood testing, HRD and hereditary testing, with each test tied to a stated case type, from newly diagnosed and locally advanced disease to late stage, rare and unknown primary cancers. Laboratories operate in India, the United Arab Emirates, Nepal and the Philippines.

The whole exome assay was validated across 17 cancer types, and the company's population studies cover 30 cancer types in Indian patients, but OncoTwin and the Cellworks ranking have no published validation in any setting.