Assessment and evaluation of heterogeneity in data from immune infiltration spatial niches in lung cancer

Abstract

The protein biomarker expressions in three types of sampled immune INFILTration spatial niches in lung cancer tissue were measured using the new
technology Digital Spatial Profiler (DSP). The three types of immune INFILTration that were observed in lung tumors were STROMA identified as
immune cells separate from tumor cells, Tertiary lymphoid structures (TLS)
identified as dense structures of organized immune cells and finally Infiltraterate where immune cells dispersed among and in direct contact with tumor
cells (INFILT). The pairwise inter and intra-patient correlation between the
protein biomarkers were evaluated using the Bland- Altman methods for
Calculating correlation coefficients with repeated observations. The result
showed that the absolute value of the inter- patient correlation levels were
higher for sample type INFILT compared to STROMA while the absolute
value of the intra- patient correlation levels were slightly higher between the
biomarkers of the sample type STROMA. In order to investigate added value
of sampling multiple regions from individual tumors, the intra- patient heterogeneity of the protein markers in the three different spatial niches were
evaluated. To this end, three different estimators were used: standard deviation, median absolute deviation and range. After comparing the results, it
was concluded that standard deviation was the preferred method. Since it
is applied on the complete set of available data and captures the behavior of
the tail of the data which is desirable for our purpose. The mean squared
error in the ANOVA table, with the patients identity as the independent
variable and marker values for each sample type as the dependent variable
was calculated as a measure of heterogeneity of the markers within patients.

Popular Abstract

In cancer treatment, drugs that modulate the immune system to more effectively combat the tumor, known as immunotherapy, have become increasingly
important. The form and level of immune infiltration in tumors vary substantially within and between tumors. A new technology make it possible
to select certain areas on the tumor tissue samples and measure the amount
of certain proteins in the chosen areas. Analysis of the protein expressions
is then the key to identify proteins that could assist us in tumor subtyping,
determination of prognosis and optimal therapeutic strategy for individual
patients. In this study, Three types of immune infiltration spatial niches in
lung cancer tissue have been sampled and more than 40 proteins have been
measured in the chosen areas of the samples.
We started our analysis of the protein expressions by examining their distribution in the three sampled regions. This approach could be helpful in
pinpointing potential targets associated with the presence of particular immune niches.
Next, we examined the linear relationship between the measured protein
pairs both between and within patients in the three sampled regions. We
had to take into consideration the fact that some samples came from the
same patients and therefore could not be assumed to be independent samples.
Establishing the strength and direction of this relationship is in particular
important for variable selection in future model building.
Since the samples come from different types of tissue structures, variation
in the measured protein expressions was expected. The challenge was how
to assess the level of variation in the measurements to potentially make use
2
of, for instance, in predicting the outcome of a certain treatment. Different
methods were used to estimates of the variability in protein values within
patients.
The result suggested an overall reduction in the magnitude of the withinpatient pairwise linear relationship of the proteins compared to the linear
relationship observed within patients. This could potentially indicate that
the relationship varies in different subgroups of the patients and needs to be
further investigated with more samples.
The result of the estimated variation in protein expressions, suggested
that level of variability varied between spatial immune infiltration niches
and between the different biomarkers. This would highlights the added value
in sampling multiple regions of different types from the same tumor. However, similarities between the results obtained using different methods could
potentially be worrisome. Further examination of the result indicated that
some of the similar patterns could be, for instance, associated with the vulnerability of the estimators to deviation of the distribution of the data from
a normal or symmetric distribution. The measurement of the variability and
its potential predictive value could be further studied in the future by employing estimators that are less sensitive to the shape of the distribution of
the data.

• author: Keivani Najafabadi, Mahta
• supervisor: Magnus Wiktorsson
• organization: Mathematical Statistics
• course: MASK11 20212
• year: 2023
• type: M2 - Bachelor Degree
• subject: Mathematics and Statistics
• publication/series: Bachelor's Theses in Mathematicas Sciences
• report number: LUNFMS-4069-2023
• ISSN: 1654-6229
• other publication id: 2023:K3
• language: English
• id: 9112848
• date added to LUP: 2023-03-24 16:25:20
• date last changed: 2023-03-27 13:15:15