Machine learning enables detection of early-stage colorectal cancer by whole-genome sequencing of plasma cell-free DNA.

BACKGROUND: Blood-based methods using cell-free DNA (cfDNA) are under development as an alternative to existing screening tests. However, early-stage detection of cancer using tumor-derived cfDNA has proven challenging because of the small proportion of cfDNA derived from tumor tissue in early-stage disease. A machine learning approach to discover signatures in cfDNA, potentially reflective of both tumor and non-tumor contributions, may represent a promising direction for the early detection of cancer. METHODS: Whole-genome sequencing was performed on cfDNA extracted from plasma samples (N = 546 colorectal cancer and 271 non-cancer controls). Reads aligning to protein-coding gene bodies were extracted, and read counts were normalized. cfDNA tumor fraction was estimated using IchorCNA. Machine learning models were trained using k-fold cross-validation and confounder-based cross-validations to assess generalization performance. RESULTS: In a colorectal cancer cohort heavily weighted towards early-stage cancer (80% stage I/II), we achieved a mean AUC of 0.92 (95% CI 0.91-0.93) with a mean sensitivity of 85% (95% CI 83-86%) at 85% specificity. Sensitivity generally increased with tumor stage and increasing tumor fraction. Stratification by age, sequencing batch, and institution demonstrated the impact of these confounders and provided a more accurate assessment of generalization performance. CONCLUSIONS: A machine learning approach using cfDNA achieved high sensitivity and specificity in a large, predominantly early-stage, colorectal cancer cohort. The possibility of systematic technical and institution-specific biases warrants similar confounder analyses in other studies. Prospective validation of this machine learning method and evaluation of a multi-analyte approach are underway.

The impact of a fourteen-gene molecular assay on physician treatment decisions in non-small-cell lung cancer.

BACKGROUND: Five-year survival in early-stage, non-squamous, non-small-cell lung cancer (NSCLC) remains poor compared with other solid tumors, even after complete resection. Post-operative management depends on prognostic staging to identify individuals at highest risk for death, and therefore with the greatest need for further intervention. A 14-gene quantitative RT-PCR test successfully differentiates stage I-III NSCLC patients who are at high-, intermediate-, or low-risk for 5-year mortality. This study assesses the impact of the assay's prognostic information on physician decisions regarding adjuvant chemotherapy. METHODS: We invited 115 physicians who ordered the test to participate in an on-line survey. The primary outcome measure was the proportion of patients with different pre- and post-test chemotherapy recommendations. RESULTS: Fifty-eight physicians (50 %) completed the survey on 120 stage I or II NSCLC patients. Ninety-one patients (76 %) had stage I lung cancer; 27 (23 %), 39 (33 %), and 54 (45 %) patients had low-, intermediate-, and high-risk scores, respectively. Physicians' chemotherapy recommendations were changed post-testing in 37 patients (30.8 %, 95 % CI 22.7-39.9 %). High-risk patients were more likely to have a change in treatment recommendation (44.4 %, 95 % CI 30.9-58.6 %) than low risk patients (3.7 %, 95 % CI 0.1-19.0 %); a substantial number of changes were observed in both stage I (33.0 %, 95 % CI 23.5-43.6 %) and stage II (24.1 %, 95 % CI 10.3-43.5 %). CONCLUSIONS: Our data show that the assay resulted in a significant impact on physician treatment decisions in early-stage NSCLC, and that the nature of treatment changes generally correlated with the test's assessment of risk.

Differential impact of test performance characteristics on burden-to-benefit tradeoffs for blood-based colorectal cancer screening: A microsimulation analysis.

OBJECTIVES: To inform the development and evaluation of new blood-based colorectal cancer (CRC) screening tests satisfying minimum United States (US) coverage criteria, we estimated the impact of the different test performance characteristics on long-term testing benefits and burdens. METHODS: A novel CRC-Microsimulation of Adenoma Progression and Screening (CRC-MAPS) model was developed, validated, then used to assess different screening tests for CRC. We compared multiple, hypothetical blood-based CRC screening tests satisfying minimum coverage criteria of 74% CRC sensitivity and 90% specificity, to measure how changes in a test's CRC sensitivity, specificity, and adenoma sensitivity (sizes 1-5 mm, 6-9 mm, ≥10 mm) affect total number of colonoscopies (COL), CRC incidence reduction (IR), CRC mortality reduction (MR), and burden-to-benefit ratios (incremental COLs per percentage-point increase in IR or MR). RESULTS: A blood test meeting minimum US coverage criteria for performance characteristics resulted in 1576 lifetime COLs per 1000 individuals, 46.7% IR and 59.2% MR compared to no screening. Tests with increased CRC sensitivity of 99% ( + 25%) vs. increased ≥10 mm adenoma sensitivity of 13.6% ( + 3.6%) both yielded the same MR, 62.7%. Test benefits improved the most with increases in all-size adenoma sensitivity, then size-specific adenoma sensitivities, then specificity and CRC sensitivity, while increases in specificity or ≥10 mm adenoma sensitivity resulted in the most favorable burden-to-benefit tradeoffs (ratios <11.5). CONCLUSIONS: Burden-to-benefit ratios for blood-based CRC screening tests differ by performance characteristic, with the most favorable tradeoffs resulting from improvements in specificity and ≥10 mm adenoma sensitivity.

Intrinsic and extrinsic pathway signaling during neuronal apoptosis: lessons from the analysis of mutant mice.

Trophic factor deprivation (TFD)-induced apoptosis in sympathetic neurons requires macromolecular synthesis-dependent BAX translocation, cytochrome c (cyt c) release, and caspase activation. Here, we report the contributions of other intrinsic and extrinsic pathway signals to these processes. Sympathetic neurons expressed all antiapoptotic BCL-2 proteins examined, yet expressed only certain BH3-only and multidomain proapoptotic BCL-2 family members. All coexpressed proapoptotic proteins did not, however, exhibit functional redundancy or compensatory expression, at least in the Bax-/-, Bak-/-, Bim-/-, Bid-/-, and Bad-/- neurons examined. Although the subcellular distribution or posttranslational modification of certain BCL-2 proteins changed with TFD, neither transcriptional nor posttranslational mechanisms regulated the expression or subcellular localization of BID, BAD, or BAK in this paradigm. Despite modest induction of Fas and FasL expression, Fas-mediated signaling did not contribute to TFD-induced apoptosis in sympathetic neurons. Similar findings were obtained with K+ withdrawal-induced apoptosis in cerebellar granule neurons, a model for activity-dependent neuronal survival in the CNS. Thus, expression alone does not guarantee functional redundancy (or compensation) among BCL-2 family members, and, at least in some cells, extrinsic pathway signaling and certain BH3-only proteins (i.e., BID and BAD) do not contribute to BAX-dependent cyt c release or apoptosis caused by TFD.

JNK-mediated BIM phosphorylation potentiates BAX-dependent apoptosis.

Trophic factor deprivation (TFD) activates c-Jun N-terminal kinases (JNKs), culminating in coordinate AP1-dependent transactivation of the BH3-only BCL-2 proteins BIM(EL) and HRK, which in turn are critical for BAX-dependent cytochrome c release, caspase activation, and apoptosis. Here, we report that TFD caused not only induction but also phosphorylation of BIM(EL). Mitochondrially localized JNKs but not upstream activators, like mixed-lineage kinases (MLKs) or mitogen-activated protein kinase kinases (MKKs), specifically phosphorylated BIM(EL) at Ser65, potentiating its proapoptotic activity. Inhibition of the JNK pathway attenuated BIM(EL) expression, prevented BIM(EL) phosphorylation, and abrogated TFD-induced apoptosis. Conversely, activation of this pathway promoted BIM(EL) expression and phosphorylation, causing BIM- and BAX-dependent cell death. Thus, JNKs regulate the proapoptotic activity of BIM(EL) during TFD, both transcriptionally and posttranslationally.

Microsatellite instability and mismatch repair protein defects in ovarian epithelial neoplasms in patients 50 years of age and younger.

Ovarian malignancies occurring in the setting of hereditary nonpolyposis colorectal carcinoma syndrome typically present in young women, often as the first or "sentinel" cancer, but the frequency of microsatellite instability (MSI) and mismatch repair (MMR) defects in ovarian surface epithelial malignancies in women

A multicenter study of the frequency and distribution of GJB2 and GJB6 mutations in a large North American cohort.

PURPOSE: The aim of the study was to determine the actual GJB2 and GJB6 mutation frequencies in North America after several years of generalized testing for autosomal recessive nonsyndromic sensorineural hearing loss to help guide diagnostic testing algorithms, especially in light of molecular diagnostic follow-up to universal newborn hearing screening. METHODS: Mutation types, frequencies, ethnic distributions, and genotype-phenotype correlations for GJB2 and GJB6 were assessed in a very large North American cohort. RESULTS: GJB2 variants were identified in 1796 (24.3%) of the 7401 individuals examined, with 399 (5.4%) homozygous and 429 (5.8%) compound heterozygous. GJB6 deletion testing was performed in 12.0% (888/7401) of all cases. The >300-kb deletion was identified in only nine individuals (1.0%), all of whom were compound heterozygous for mutations in GJB2 and GJB6. Among a total of 139 GJB2 variants identified, 53 (38.1%) were previously unreported, presumably representing novel pathogenic or benign variants. CONCLUSIONS: The frequency and distribution of sequence changes in GJB2 and GJB6 in North America differ from those previously reported, suggesting a considerable role for loci other than GJB2 and GJB6 in the etiology of autosomal recessive nonsyndromic sensorineural hearing loss, with minimal prevalence of the GJB6 deletion.