Genetix: A Deterministic-Bayesian and Generative AI Framework for Mendelian and Polygenic Trait Inheritance Prediction

Genetics education and genetic counselling have long relied on two separate kinds of software: simple rule-based calculators that are accurate but only work for single-gene traits, and machine-learning trait predictors that generalize well but cannot explain their own reasoning. Genetix is an open-source, browser-based platform built to combine the strengths of both. It uses an exact Bayesian network to compute single-gene (Mendelian) inheritance probabilities — such as blood type, Rh factor, and X-linked conditions — with guaranteed logical accuracy, and a Variational Autoencoder (VAE) to estimate probability distributions for complex, multi-gene (polygenic) traits such as height. A generative-AI layer (Gemini API) turns these computed results into plain-language explanations but is deliberately never allowed to invent a probability of its own. We describe this dual-engine architecture, ground its design decisions in published genetics and obstetric literature, and validate its behaviour across seven test scenarios covering blood-type incompatibility, high-risk pregnancy flagging, classical Mendelian ratios, X-linked inheritance, pedigree export, and polygenic height estimation. All deterministic scenarios matched the expected outcome on every repeated run, and the polygenic module produced a bounded distribution centred near the expected value rather than a single, overconfident number. We conclude that keeping exact genetic logic and learned statistical models in clearly separated layers — with generative AI confined to a narration role — is a practical and defensible pattern for genetics software that must be both rigorous and accessible.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-18
DOI
https://doi.org/10.5281/zenodo.24340694
Primary Topic
Artificial Intelligence in Healthcare and Education
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Genetix: A Deterministic-Bayesian and Generative AI Framework for Mendelian and Polygenic Trait Inheritance Prediction

Pratik Thakare, Mohan Yadav, Pooja Chauhan, Sagar Tayade
3 links
Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence in Healthcare and Education
preprint

Genetix: A Deterministic-Bayesian and Generative AI Framework for Mendelian and Polygenic Trait Inheritance Prediction

Pratik Thakare, Mohan Yadav, Pooja Chauhan, Sagar Tayade
preprint en

Abstract

Genetics education and genetic counselling have long relied on two separate kinds of software: simple rule-based calculators that are accurate but only work for single-gene traits, and machine-learning trait predictors that generalize well but cannot explain their own reasoning. Genetix is an open-source, browser-based platform built to combine the strengths of both. It uses an exact Bayesian network to compute single-gene (Mendelian) inheritance probabilities — such as blood type, Rh factor, and X-linked conditions — with guaranteed logical accuracy, and a Variational Autoencoder (VAE) to estimate probability distributions for complex, multi-gene (polygenic) traits such as height. A generative-AI layer (Gemini API) turns these computed results into plain-language explanations but is deliberately never allowed to invent a probability of its own. We describe this dual-engine architecture, ground its design decisions in published genetics and obstetric literature, and validate its behaviour across seven test scenarios covering blood-type incompatibility, high-risk pregnancy flagging, classical Mendelian ratios, X-linked inheritance, pedigree export, and polygenic height estimation. All deterministic scenarios matched the expected outcome on every repeated run, and the polygenic module produced a bounded distribution centred near the expected value rather than a single, overconfident number. We conclude that keeping exact genetic logic and learned statistical models in clearly separated layers — with generative AI confined to a narration role — is a practical and defensible pattern for genetics software that must be both rigorous and accessible.

Zenodo (CERN European Organization for Nuclear Research)
Thakur College of Engineering and Technology (IN)
Artificial Intelligence in Healthcare and Education
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon