Last week PhD student Bilge Sari published her first paper in Frontiers in Cell and Molecular Biology! It’s a really nice body of work where She and Ozcan Gulbey analysed loads of data related to pancreatic adenocarcinoma and identified that the expression of a specific laminin, laminin 332, can be used as a biomarker to predict patient outcomes. Really interesting, valuable work.
This study was possible due to one of the great things about modern science; data availability. A condition of publishing nowadays is that the underlying data from the study becomes publically available. In practice this means that science becomes a global endeavour working toward common goals. It allows big data sets to be combined to ask complex questions with sufficient sample sizes to make robust inferences.
Bilge’s work really harnessed this big data power. She took essentially everything that is out there relating to pancreatic adenocarcinoma and combined those data into a complete narrative. Specifically she looked at mRNA expression in multiple modalities, genetic changes, protein and signalling pathway changes, drug responses. In each case examining the laminin family and their association with disease states and clinical features. It is a massive body of work.
Why laminins?
If you have ever been in a meeting with any member of my team, you’ll know that laminins (link to friendly intro) are truly amazing proteins with important roles in a wide array of health and disease.
They have been studied in pancreatic cancer before, with individual studies showing correlations between single family members and outcomes and funcitonal studies showing that those changes affect cancer cell behaviour.
What we felt was missing, was taking into account that the each mature laminin is actually made up of three individual subunits. Therefore, when considering outcomes, one should look at the family rather than a single individual chain.
In Bilge’s analyses, it became quickly clear from patient data that high expression of three of the laminin genes, LAMA3, LAMB3 or LAMC2 were associated with worse outcomes. However, the predictive value and effect size got even larger when two or three of these genes were combined.In the two figures below you can see this in action.
The first compares the low expression group and high expression group for the three-gene signature. As you can see, more than 75% of the high expression group of patients died from the disease after 4 years compared with only about 30% of the low expression group.
We talk about these differences in patients as hazard ratios, with a HR of 1 meaning no difference between the groups. Here the HR was 3.4, the high expression group were 3.4 times more likely to die than the low expression group.
The second figure focuses more on these hazard ratios (dots), with confidence intervals indicated by the lines. Any time where the confidence interval lines overlap with the red dotted line at 1, means we can’t be confident there is a difference between the groups. As you can see LAMA3, LAMB3 and LAMC2 are shifted to the right and don’t overlap, meaninbg higher risk and also greater confidence. All the ones at the bottom, where genes are combined, shift the dot and lines further to the right.
These data are just the tip of the iceberg in terms of this paper (these are panels A and I in Figure 5 and there are 9 figures + 10 supplemental figs!). I encourage you to explore more.
As usual you can read or download the paper for free via the button below.
LaNts and Laminins 