Protein design and variant prediction using autoregressive generative models.

The ability to design functional sequences and predict effects of variation is central to protein engineering and biotherapeutics. State-of-art computational methods rely on models that leverage evolutionary information but are inadequate for important applications where multiple sequence alignments are not robust. Such applications include the prediction of variant effects of indels, disordered proteins, and the design of proteins such as antibodies due to the highly variable complementarity determining regions. We introduce a deep generative model adapted from natural language processing for prediction and design of diverse functional sequences without the need for alignments. The model performs state-of-art prediction of missense and indel effects and we successfully design and test a diverse 105-nanobody library that shows better expression than a 1000-fold larger synthetic library. Our results demonstrate the power of the alignment-free autoregressive model in generalizing to regions of sequence space traditionally considered beyond the reach of prediction and design.

Non coding RNA analysis in fibrolamellar hepatocellular carcinoma.

Fibrolamellar hepatocellular carcinoma (FLC) is a rare primary liver cancer found in adolescents and young adults without underlying liver disease. A deletion of ~400 kD has been found in one copy of chromosome 19 in the tumor tissue of all patients tested. This produces a fusion of the genes DNAJB1 and PRKACA which, in turn, produces a chimeric transcript and protein. Transcriptomic analysis of the tumor has shown upregulation of various oncologically relevant pathways, including EGF/ErbB, Aurora Kinase A, pak21 and wnt. To explore other factors that may contribute to oncogenesis, we examined the microRNA (miRNA) and long non-coding RNA (lncRNA) expression in FLC. The non-coding RNA expression profile in tumor tissue samples is distinctly different from the adjacent normal liver and from other liver tumors. Furthermore, miRZip knock down or over expression of certain miRNAs led to changes in the levels of coding genes that recapitulated changes observed in FLC, suggesting mechanistically that the changes in the cellular levels of miRNA are not merely correlative. Thus, in addition to serving as diagnostic tools for FLC, non-coding RNAs may serve as therapeutic targets.

Transcriptomic characterization of fibrolamellar hepatocellular carcinoma.

Fibrolamellar hepatocellular carcinoma (FLHCC) tumors all carry a deletion of ∼ 400 kb in chromosome 19, resulting in a fusion of the genes for the heat shock protein, DNAJ (Hsp40) homolog, subfamily B, member 1, DNAJB1, and the catalytic subunit of protein kinase A, PRKACA. The resulting chimeric transcript produces a fusion protein that retains kinase activity. No other recurrent genomic alterations have been identified. Here we characterize the molecular pathogenesis of FLHCC with transcriptome sequencing (RNA sequencing). Differential expression (tumor vs. adjacent normal tissue) was detected for more than 3,500 genes (log2 fold change ≥ 1, false discovery rate ≤ 0.01), many of which were distinct from those found in hepatocellular carcinoma. Expression of several known oncogenes, such as ErbB2 and Aurora Kinase A, was increased in tumor samples. These and other dysregulated genes may serve as potential targets for therapeutic intervention.

Detection of a recurrent DNAJB1-PRKACA chimeric transcript in fibrolamellar hepatocellular carcinoma.

Fibrolamellar hepatocellular carcinoma (FL-HCC) is a rare liver tumor affecting adolescents and young adults with no history of primary liver disease or cirrhosis. We identified a chimeric transcript that is expressed in FL-HCC but not in adjacent normal liver and that arises as the result of a ~400-kilobase deletion on chromosome 19. The chimeric RNA is predicted to code for a protein containing the amino-terminal domain of DNAJB1, a homolog of the molecular chaperone DNAJ, fused in frame with PRKACA, the catalytic domain of protein kinase A. Immunoprecipitation and Western blot analyses confirmed that the chimeric protein is expressed in tumor tissue, and a cell culture assay indicated that it retains kinase activity. Evidence supporting the presence of the DNAJB1-PRKACA chimeric transcript in 100% of the FL-HCCs examined (15/15) suggests that this genetic alteration contributes to tumor pathogenesis.