A novel machine learning algorithm may help some clinicians to determine whether a patient has inherited or immune/acquired bone marrow failure (BMF), according to research published in Blood. While the tool’s sensitivity was only approximately 89%, it may help non-specialists to better streamline patients for confirmatory genetic testing.

A major difficulty in treating BMF is that therapeutic decision-making depends on distinguishing the syndrome’s etiology — the same treatment for inherited BMF is not recommended for acquired BMF. Moreover, genetic testing is not always available, particularly in the low-resource setting.

The potential for misdiagnosis and incorrect treatment makes novel tools for determining BMF etiology essential. For this study, researchers trained and tested a machine learning algorithm that aimed to distinguish inherited and acquired BMF using existing patient data.

Continue Reading

Overall, data from 359 and 127 patients were used in the algorithm’s training and validation datasets, respectively. In the training and validation sets, 64.6% and 72.4% of patients had acquired BMF, respectively, 35.3% and 27.5% had inherited BMF, the median ages were 28 and 23 years, and 48% and 46% of patients were female sex.

The researchers created a model using 25 variables frequently recorded at a patient’s first clinical encounter. An algorithm unbiasedly clustered the cases into 1 of 2 groups: Cluster A, which was mostly comprised of immune or aplastic anemia, and Cluster B, which was comprised of a small group of underrepresented BMF phenotypes. The latter cluster was not, however, included in subsequent analysis because of its small size.

Analysis of the model’s assignments showed that the algorithm was accurate in 89% of cases in predicting BMF etiology. Specifically, the model correctly predicted inherited BMF in 79% of cases and likely immune in 92% of cases.

“This practical tool is also part of ongoing research because we will continue accruing to the model in an effort to increase the number of cases to further refine prediction of [inherited BMF syndrome] cases that were underrepresented in the current cohort, especially pediatric cases,” the authors noted in their report.

Disclosures: Some study authors declared affiliations with biotech, pharmaceutical, or device companies. Please see the original reference for a full list of disclosures.


Gutierrez-Rodrigues F, Munger E, Ma X, et al. Differential diagnosis of bone marrow failure syndromes guided by machine learning. Blood. 2023;141(17):2100-2113. doi:10.1182/blood.2022017518