Dataset Information

A diagnostic classifier for gene expression-based identification of early Lyme disease.

ABSTRACT:

Background

Lyme disease is a tick-borne illness that causes an estimated 476,000 infections annually in the United States. New diagnostic tests are urgently needed, as existing antibody-based assays lack sufficient sensitivity and specificity.

Methods

Here we perform transcriptome profiling by RNA sequencing (RNA-Seq), targeted RNA-Seq, and/or machine learning-based classification of 263 peripheral blood mononuclear cell samples from 218 subjects, including 94 early Lyme disease patients, 48 uninfected control subjects, and 57 patients with other infections (influenza, bacteremia, or tuberculosis). Differentially expressed genes among the 25,278 in the reference database are selected based on ≥1.5-fold change, ≤0.05 p value, and ≤0.001 false-discovery rate cutoffs. After gene selection using a k-nearest neighbor algorithm, the comparative performance of ten different classifier models is evaluated using machine learning.

Results

We identify a 31-gene Lyme disease classifier (LDC) panel that can discriminate between early Lyme patients and controls, with 23 genes (74.2%) that have previously been described in association with clinical investigations of Lyme disease patients or in vitro cell culture and rodent studies of Borrelia burgdorferi infection. Evaluation of the LDC using an independent test set of samples from 63 subjects yields an overall sensitivity of 90.0%, specificity of 100%, and accuracy of 95.2%. The LDC test is positive in 85.7% of seronegative patients and found to persist for ≥3 weeks in 9 of 12 (75%) patients.

Conclusions

These results highlight the potential clinical utility of a gene expression classifier for diagnosis of early Lyme disease, including in patients negative by conventional serologic testing.

SUBMITTER: Servellita V

PROVIDER: S-EPMC9306241 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:A bioinformatics approach was employed to identify transcriptome alterations in the peripheral blood mononuclear cells of well-characterized human subjects who were diagnosed with early disseminated Lyme disease (LD) based on stringent microbiological and clinical criteria. Transcriptomes were assessed at the time of presentation and also at approximately 1 month (early convalescence) and 6 months (late convalescence) after initiation of an appropriate antibiotic regimen. Comparative transcriptomics identified 335 transcripts, representing 233 unique genes, with significant alterations of at least 2-fold expression in acute- or convalescent-phase blood samples from LD subjects relative to healthy donors. Acute-phase blood samples from LD subjects had the largest number of differentially expressed transcripts (187 induced, 54 repressed). This transcriptional profile, which was dominated by interferon-regulated genes, was sustained during early convalescence. 6 months after antibiotic treatment the transcriptome of LD subjects was indistinguishable from that of healthy controls based on two separate methods of analysis. Return of the LD expression profile to levels found in control subjects was concordant with disease outcome; 82% of subjects with LD experienced at least one symptom at the baseline visit compared to 43% at the early convalescence time point and only a single patient (9%) at the 6-month convalescence time point. Using the random forest machine learning algorithm, we developed an efficient computational framework to identify sets of 20 classifier genes that discriminated LD from other bacterial and viral infections. These novel LD biomarkers not only differentiated subjects with acute disseminated LD from healthy controls with 96% accuracy but also distinguished between subjects with acute and resolved (late convalescent) disease with 97% accuracy.IMPORTANCE Lyme disease (LD), caused by Borrelia burgdorferi, is the most common tick-borne infectious disease in the United States. We examined gene expression patterns in the blood of individuals with early disseminated LD at the time of diagnosis (acute) and also at approximately 1 month and 6 months following antibiotic treatment. A distinct acute LD profile was observed that was sustained during early convalescence (1 month) but returned to control levels 6 months after treatment. Using a computer learning algorithm, we identified sets of 20 classifier genes that discriminate LD from other bacterial and viral infections. In addition, these novel LD biomarkers are highly accurate in distinguishing patients with acute LD from healthy subjects and in discriminating between individuals with active and resolved infection. This computational approach offers the potential for more accurate diagnosis of early disseminated Lyme disease. It may also allow improved monitoring of treatment efficacy and disease resolution.

Project description:Although osteosarcoma (OS) is a rare cancer, it is the most common primary malignant bone tumor in children and adolescents. BRCAness is a phenotypical trait in tumors with a defect in homologous recombination repair, resembling tumors with inactivation of BRCA1/2, rendering these tumors sensitive to poly (ADP)-ribose polymerase inhibitors (PARPi). Recently, OS was shown to exhibit molecular features of BRCAness. Our goal was to develop a method complementing existing genomic methods to aid clinical decision making on administering PARPi in OS patients. OS samples with DNA-methylation data were divided to BRCAness-positive and negative groups based on the degree of their genomic instability (n = 41). Methylation probes were ranked according to decreasing variance difference between two groups. The top 2000 probes were selected for training and cross-validation of the random forest algorithm. Two-thirds of available OS RNA-Seq samples (n = 17) from the top and bottom of the sample list ranked according to genome instability score were subjected to differential expression and, subsequently, to gene set enrichment analysis (GSEA). The combined accuracy of trained random forest was 85% and the average area under the ROC curve (AUC) was 0.95. There were 449 upregulated and 1,079 downregulated genes in the BRCAness-positive group (fdr < 0.05). GSEA of upregulated genes detected enrichment of DNA replication and mismatch repair and homologous recombination signatures (FWER < 0.05). Validation of the BRCAness classifier with an independent OS set (n = 20) collected later in the course of study showed AUC of 0.87 with an accuracy of 90%. GSEA signatures computed for this test set were matching the ones observed in the training set enrichment analysis. In conclusion, we developed a new classifier based on DNA-methylation patterns that detects BRCAness in OS samples with high accuracy. GSEA identified genome instability signatures. Machine-learning and gene expression approaches add new epigenomic and transcriptomic aspects to already established genomic methods for evaluation of BRCAness in osteosarcoma and can be extended to cancers characterized by genome instability.