An integrated molecular diagnostic report for heart transplant biopsies using an ensemble of diagnostic algorithms

https://doi.org/10.1016/j.healun.2019.01.1318Get rights and content

BACKGROUND

We previously reported a microarray-based diagnostic system for heart transplant endomyocardial biopsies (EMBs), using either 3-archetype (3AA) or 4-archetype (4AA) unsupervised algorithms to estimate rejection. In the present study we examined the stability of machine-learning algorithms in new biopsies, compared 3AA vs 4AA algorithms, assessed supervised binary classifiers trained on histologic or molecular diagnoses, created a report combining many scores into an ensemble of estimates, and examined possible automated sign-outs.

METHODS

We studied 889 EMBs from 454 transplant recipients at 8 centers: the initial cohort (N = 331) and a new cohort (N = 558). Published 3AA algorithms derived in Cohort 331 were tested in Cohort 558, the 3AA and 4AA models were compared, and supervised binary classifiers were created.

RESULTS

A`lgorithms derived in Cohort 331 performed similarly in new biopsies despite differences in case mix. In the combined cohort, the 4AA model, including a parenchymal injury score, retained correlations with histologic rejection and DSA similar to the 3AA model. Supervised molecular classifiers predicted molecular rejection (areas under the curve [AUCs] >0.87) better than histologic rejection (AUCs <0.78), even when trained on histology diagnoses. A report incorporating many AA and binary classifier scores interpreted by 1 expert showed highly significant agreement with histology (p < 0.001), but with many discrepancies, as expected from the known noise in histology. An automated random forest score closely predicted expert diagnoses, confirming potential for automated signouts.

CONCLUSIONS

Molecular algorithms are stable in new populations and can be assembled into an ensemble that combines many supervised and unsupervised estimates of the molecular disease states.

Section snippets

Population

This prospective study was approved by the ethics review board of each center (see Table S1 online) and is registered at ClinicalTrials.gov (NCT02670408). As described elsewhere,12 biopsies were collected prospectively for clinical indications or by protocol, and then processed for histology and human leukocyte antibody (HLA) testing as per local standard of care. Histology interpretation followed ISHLT guidelines.3, 22

Microarray analysis

As detailed elsewhere,12 purified total RNA from 889 EMBs, which included

Biopsy population

Clinically indicated, protocol, and follow-up EMBs from 454 patients were collected in 2 cohorts: cohort 331 (N = 331) from the earlier study12 and a new cohort of 558 biopsies (Table 1, and Table S2 online). Cardiomyopathy and coronary artery disease were the common primary diseases.

Histology grades and molecular scores were translated to a common nomenclature13 (see Methods): TCMR, pTCMR, ABMR, and pABMR. Frequencies of the histologic classes in Cohorts 331 and 558 are presented in Table 2

Discussion

Having found value in molecular EMB assessment (especially after incorporating injury estimates),13 in the present study we aimed to develop an ensemble of estimates into a biopsy report and examine automated sign-out. We established that machine-learning algorithms trained in one cohort perform similarly in future biopsies despite differences in case mix. We showed that Model 2 incorporating injury demonstrated similar relationships to histologic rejection as Model 1 and correlated with DSA

Disclosure statement

P.F.H. holds shares in Transcriptome Sciences, Inc., a University of Alberta research company with an interest in molecular diagnostics, and has been a symposium speaker for One Lambda. The other authors no conflicts of interest to disclose.

This study was supported by funds and/or resources from Genome Canada, the Canada Foundation for Innovation, the University of Alberta Hospital Foundation, the Alberta Ministry of Advanced Education and Technology, the Mendez National Institute of

Supplementary data

Supplementary data associated with this article can be found in the online version at www.jhltonline.org/.

References (34)

Cited by (40)

  • The real rejection stands up and displays its complexities

    2022, Journal of Heart and Lung Transplantation
View all citing articles on Scopus
View full text