Irregulab: Automated interpretation of irregular antibody panels

Abstract

Abstract Introduction Identifying irregular red-cell antibodies is an essential but still largely manual task in transfusion medicine. Interpreting antibody panels means weighing complex serologic patterns while accounting for dosage effects, enzyme effects, and cold reactivity. Manual interpretation is time-intensive and prone to inter-observer variation, especially with multiple or weak reactions. IrregulAB aims to aid interpretation and enhance consistency of this process across laboratories. Objectives Develop and validate IrregulAB, an open-source, rule-based engine that automates red blood cell antibody panel interpretation by replicating expert reasoning, reducing interpretation time, increasing consistency, and providing a transparent and reproducible solution for routine laboratory use. Methodology IrregulAB is an open-source, rule-based engine that emulates expert reasoning when interpreting red-cell antibody panels. The algorithm proceeds in two stages. Stage 1 – Antibody triage. Results from low-ionic-strength saline (LISS), enzyme-treated, and 4 °C phases are analyzed sequentially. For each antibody, the engine assigns one of three possible statuses based on the phase-by-phase analysis: retained (if compatible with the panel data), discarded (if clearly incompatible), or flagged (if results suggest possible potentiation or other uncertainties). Ubiquitous or untested antigens are ignored as they provide no discriminatory power. Dosage effects are considered: heterozygous antigen expression may weaken reactions, so an antibody present in a non-reactive cell is discarded except when dosage, enzyme destruction/enhancement, or cold potentiation can account for the discrepancy. Stage 2 – Minimal-set search. Using a conservative OR rule—a cell is positive if any retained antibody targets one of its antigens—the engine enumerates the smallest antibody combinations that reproduce the panel results. Rows containing only heterozygous antigens are ignored. Validation. We evaluated IrregulAB on 116 routine antibody panels (2024) and an external set of 38 panels (March–June 2025), all performed on Bio-Rad ID-DiaMed gel cards by the Blood Bank of the Santiago de Compostela Clinic Hospital. Ground-truth identifications were provided by senior blood-bank physicians following standard protocols. Results To evaluate the performance of the system, we conducted validation on an independent test set that was not involved in the algorithm design. IrregulAB matched expert calls in 93 % of retained antibodies and 87 % of minimal antibody sets (91 % / 89 % on the external set). The 18 discordant panels were mainly attributable to antigen-dosage weak P1/M reactions (8), phase-wide panagglutination (3), atypical enzyme or cold potentiation (2), early anti-C emergence later confirmed manually (2), one spurious weak cell (1), and overlap with low-prevalence anti-Cw (2). Conclusion IrregulAB matched expert interpretations in ≈ 91 % of panels, providing a transparent, reproducible decision-support aid for routine blood bank analysis. Its LGPL-licensed source code and public web interface (https://irregulab.codigorojo.tech/) enable easy adoption and peer scrutiny. With a residual false-negative rate of about 7 %, IrregulAB is intended to augment, not replace, expert oversight, shortening interpretation and harmonizing results across laboratories. Ongoing work includes multicentre validation, and expansion of the rule base to cover rare antibodies.