Improving lung cancer risk stratification leveraging whole transcriptome RNA sequencing and machine learning across multiple cohorts.

BACKGROUND: Bronchoscopy for suspected lung cancer has low diagnostic sensitivity, rendering many inconclusive results. The Bronchial Genomic Classifier (BGC) was developed to help with patient management by identifying those with low risk of lung cancer when bronchoscopy is inconclusive. The BGC was trained and validated on patients in the Airway Epithelial Gene Expression in the Diagnosis of Lung Cancer (AEGIS) trials. A modern patient cohort, the BGC Registry, showed differences in key clinical factors from the AEGIS cohorts, with less smoking history, smaller nodules and older age. Additionally, we discovered interfering factors (inhaled medication and sample collection timing) that impacted gene expressions and potentially disguised genomic cancer signals. METHODS: In this study, we leveraged multiple cohorts and next generation sequencing technology to develop a robust Genomic Sequencing Classifier (GSC). To address demographic composition shift and interfering factors, we synergized three algorithmic strategies: 1) ensemble of clinical dominant and genomic dominant models; 2) development of hierarchical regression models where the main effects from clinical variables were regressed out prior to the genomic impact being fitted in the model; and 3) targeted placement of genomic and clinical interaction terms to stabilize the effect of interfering factors. The final GSC model uses 1232 genes and four clinical covariates - age, pack-years, inhaled medication use, and specimen collection timing. RESULTS: In the validation set (N = 412), the GSC down-classified low and intermediate pre-test risk subjects to very low and low post-test risk with a specificity of 45% (95% CI 37-53%) and a sensitivity of 91% (95%CI 81-97%), resulting in a negative predictive value of 95% (95% CI 89-98%). Twelve percent of intermediate pre-test risk subjects were up-classified to high post-test risk with a positive predictive value of 65% (95%CI 44-82%), and 27% of high pre-test risk subjects were up-classified to very high post-test risk with a positive predictive value of 91% (95% CI 78-97%). CONCLUSIONS: The GSC overcame the impact of interfering factors and achieved consistent performance across multiple cohorts. It demonstrated diagnostic accuracy in both down- and up-classification of cancer risk, providing physicians actionable information for many patients with inconclusive bronchoscopy.

hATpin, a family of MITE-like hAT mobile elements conserved in diverse plant species that forms highly stable secondary structures.

We identified a 178 bp mobile DNA element in lettuce with characteristic CGAGC/GCTCG repeats in the subterminal regions. This element has terminal inverted repeats and 8-bp target site duplications typical of the hAT superfamily of class II mobile elements, but its small size and potential to form a single-stranded stable hairpin-like secondary structure suggest that it is related to MITE elements. In silico searches for related elements identified 252 plant sequences with 8-bp target site duplications and sequence similarity in their terminal and subterminal regions. Some of these sequences were predicted to encode transposases and may be autonomous elements; these constituted a separate clade within the phylogram of hAT transposases. We demonstrate that the CGAGC/GCTCG pentamer maximizes the hairpin stability compared to any other pentamer with the same C + G content, and the secondary structures of these elements are more stable than for most MITEs. We named these elements collectively as hATpin elements because of the hAT similarity and their hairpin structures. The nearly complete rice genome sequence and the highly advanced genome annotation allowed us to localize most rice elements and to deduce insertion preferences. hATpin elements are distributed on all chromosomes, but with significant bias for chromosomes 1 and 10 and in regions of moderate gene density. This family of class II mobile elements is found primarily in monocot species, but is also present in dicot species.

Analysis of the transcriptional complexity of Arabidopsis thaliana by massively parallel signature sequencing.

Large-scale sequencing of short mRNA-derived tags can establish the qualitative and quantitative characteristics of a complex transcriptome. We sequenced 12,304,362 tags from five diverse libraries of Arabidopsis thaliana using massively parallel signature sequencing (MPSS). A total of 48,572 distinct signatures, each representing a different transcript, were expressed at significant levels. These signatures were compared to the annotation of the A. thaliana genomic sequence; in the five libraries, this comparison yielded between 17,353 and 18,361 genes with sense expression, and between 5,487 and 8,729 genes with antisense expression. An additional 6,691 MPSS signatures mapped to unannotated regions of the genome. Expression was demonstrated for 1,168 genes for which expression data were previously unknown. Alternative polyadenylation was observed for more than 25% of A. thaliana genes transcribed in these libraries. The MPSS expression data suggest that the A. thaliana transcriptome is complex and contains many as-yet uncharacterized variants of normal coding transcripts.

OptiRNAi, an RNAi design tool.

RNA interference (RNAi), a recently developed reverse genetics tool, has many advantages compared to traditional gene knockout methods. Appropriate selection of double stranded RNAs identical to a specific region(s) of the target gene is critical for the successful implementation of this technology. Recently, Elbashir et al. [Methods 26 (2002) 199] has established empirical criteria for siRNA sequence selection that significantly improved the success rate for RNAi attempts. We have developed OptiRNAi, a computational tool, which uses the Elbashir et al. criteria to predict appropriate target sequences for siRNA production. Specificity of these siRNAs for the target of interest can then be assessed by the investigator using the embedded Blast search engine optimized for RNAi design. Thus, OptiRNAi is an efficient and user friendly tool for RNAi design based on criteria that are more stringent than other available tools.

Arbuscular mycorrhizal fungi enhance aluminium resistance of broomsedge (Andropogon virginicus L.).

In the eastern United States, broomsedge (Andropogon virginicus L.) is found growing on abandoned coal-mined lands that have extremely acidic soils with high residual aluminium (Al) concentrations. Broomsedge may be inherently metal-resistant and nutrient-efficient or may rely on the arbuscular mycorrhizal (AM) fungal association to overcome limitations on such sites. Broomsedge plants were grown with and without an acidic ecotype AM fungal consortium and exposed to controlled levels of Al in two experiments. The AM fungal consortium conferred Al resistance to broomsedge. Arbuscular mycorrhizal fungi reduced Al uptake and translocation in host plants, potentially reflecting measured reductions in inorganic Al availability in the rhizosphere of mycorrhizal plants. Mycorrhizal plants exhibited lower shoot P concentrations, higher phosphorus use efficiency, and lower root acid phosphatase rates than non-mycorrhizal plants. Aluminium significantly reduced calcium (Ca) and magnesium (Mg) tissue concentrations in both mycorrhizal and non-mycorrhizal plants. However, plant response to any change in nutrient acquisition was substantially less pronounced in mycorrhizal plants. The exclusion of Al and greater stability of tissue biomass accretion-tissue nutrient relationships in mycorrhizal broomsedge plants exposed to Al may be important mechanisms that allow broomsedge to grow on unfavourable acidic soils.

Diversity in nucleotide binding site-leucine-rich repeat genes in cereals.

The diversity of the largest group of plant disease resistance genes, the nucleotide binding site-leucine-rich repeat (NBS-LRR) genes, was examined in cereals following polymerase chain reaction (PCR) cloning and database mining. NBS-LRR genes in rice are a large and diverse class with more than 600 genes, at least three to four times the complement of Arabidopsis. Most occur in small families containing one or a few cross-hybridizing members. Unlike in Arabidopsis and other dicots, the class of NBS-LRR genes coding for a Toll and mammalian interleukin-1 receptor (TIR) domain were not amplified during the evolution of the cereals. Genes coding for TIR domains are present in the rice genome, but have diverged from the NBS-LRR genes. Most cereal genes are similar in structure to the members of the non-TIR class of dicots, although many do not code for a coiled-coil domain in their amino termini. One unique class of cereal genes, with ~50 members, codes for proteins similar to the N-termini and NBS domains of resistance genes but does not code for LRR domains. The resistance gene repertoire of grasses has changed from that of dicots in their independent evolution since the two groups diverged. It is not clear whether this reflects a difference in downstream defense signaling pathways.