RESUMO

The growing use of next-generation sequencing technologies on genetic diagnosis has produced an exponential increase in the number of variants of uncertain significance (VUS). In this manuscript, we compare three machine learning methods to classify VUS as Pathogenic or No pathogenic, implementing a Random Forest (RF), a Support Vector Machine (SVM), and a Multilayer Perceptron. To train the models, we extracted high-quality variants from ClinVar that were previously classified as VUS. For each variant, we retrieved nine conservation scores, the loss-of-function tool, and allele frequencies. For the RF and SVM models, hyperparameters were tuned using cross-validation with a grid search. The three models were tested on a nonoverlapping set of variants that had been classified as VUS over the last 3 years, but had been reclassified in August 2020. The three models yielded superior accuracy on this set compared to the benchmarked tools. The RF-based model yielded the best performance across different variant types and was used to create VusPrize, an open-source software tool for prioritization of VUS. We believe that our model can improve the process of genetic diagnosis in research and clinical settings.

Assuntos

Sequenciamento de Nucleotídeos em Larga Escala , Aprendizado de Máquina , Sequenciamento de Nucleotídeos em Larga Escala/métodos , Humanos , Redes Neurais de Computação , Software , Máquina de Vetores de Suporte

RESUMO

Resumen La identificación de bacilos gramnegativos no fermentadores de la glucosa (BGNNF) es una tarea compleja y laboriosa que exige la participación de expertos. Para facilitar la toma de decisiones se desarrolló y puso a prueba un Sistema Experto (SE) con una base de conocimientos construida aplicando el algoritmo C4.5 modificado, capaz de inducir un árbol de decisión (reglas primarias) para un conjunto de géneros, y la diferenciación entre éstos (reglas complementarias) para la identificación de géneros específicos. La incertidumbre del sistema es tratada mediante el esquema de factores de certeza. En este trabajo se sometió a prueba el SE con una selección de cultivos de BGNNF de diferente origen, identificados y preservados en el Centro Venezolano de Colecciones de Microorganismos (CVCM): géneros Achromobacter, Acinetobacter, Alcaligenes, Brevundimonas, Burkholderia, Chryseobacterium, Comamonas, Delftia, Moraxella, Myroides, Ochrobactrum, Oligella, Pseudomonas, Shewanella, Sphingobacterium y Stenotrophomonas. Mediante la aplicación de 11 pruebas (características primarias) se obtuvo una aproximación entre varios de los géneros posibles. Las pruebas complementarias sugeridas (entre 1 y 9), permitieron una mayor aproximación al género posible. Los resultados muestran una coincidencia del 95.8% con los reportados por el CVCM. En base a estos resultados se estudia la ampliación de la base conocimiento para la identificación de especies de BGNNF, y de otros grupos de bacterias.

---

Abstract The identification of glucose non fermentative gram-negative bacilli (NFGNB) is a complex and laborious task. In order to facilitate genera identification an Expert System (ES) was developed applying a modified C4.5 algorithm able to induce a decision tree (primary rules) for a set of genera, and the differentiation between these (complementary rules) for the identification of specific genera- The uncertainty of the system is treated by means of a certainty factors scheme. In this work the ES was put on approval using a selection of cultures of NFGNB of different origin, identified and preserved in the Venezuelan Center for Microbial Collections (CVCM): genera Achromobacter, Acinetobacter, Alcaligenes, Brevundimonas, Burkholderia, Chryseobacterium, Comamonas, Delftia, Moraxella, Myroides, Ochrobactrum, Oligella, Pseudomonas, Shewanella, Sphingobacterium y Stenotrophomonas. By means of the application of 11 tests (primary characteristics) an approach between several of possible genera was obtained, The suggested complementary testes (between 1 to 9) allowed great approach to the possible generas.-.The results show a coincidence of the 95.8% with the reported ones by the CVCM. An extent of the ES for the identification of other genera is under study.