5  panel

The R code chunks in Chapter 5 require loading the ordPanel package,

library(ordPanel)

The example data and statistical analyses presented in Chapter 5 were previously published in a peer-reviewed manuscript by Zezulinski et al. (2025).

NoteData Example

Zezulinski D, Hoteit M, Kaplan D, Simeone A, Zhan T, Doria C, Ahmed F, Roberts L, Block T, Sayeed A (2025). “Detection of circulating mRNA variants in hepatocellular carcinoma patients using targeted RNAseq.” Liver Cancer. ISSN 2235-1795, doi:10.1159/000545366 https://doi.org/10.1159/000545366.

5.1 An S4 Class 'panel'

The function panel() (Chapter 1) creates an R object of S4 class 'panel', with slots

• @m1, a logical matrix, variant-signatures on the rows and positive subjects on the columns (Listing 5.1)
• @m0, a logical matrix, variant-signatures on the rows and negative subjects on the columns (Listing 5.2)
• @id, a list of character vectors. Each character vector contains the variants in a signature (Listing 1.7)
• @label, an optional character scalar

Listing 5.1: Slot @m1 of S4 class 'panel' (Listing 1.5) (Zezulinski et al. 2025)

m@m1 |> 
  head(n = c(3L, 3L))
#                UPH-106P UPH-004P UPH-083P
# Signature    1    FALSE    FALSE     TRUE
# Signature    2    FALSE    FALSE    FALSE
# Signature    3    FALSE    FALSE     TRUE

Listing 5.2: Slot @m0 of S4 class 'panel' (Listing 1.5) (Zezulinski et al. 2025)

m@m0 |> 
  head(n = c(3L, 3L))
#                UPC-022P UPC-021P UPC-044P
# Signature    1    FALSE    FALSE    FALSE
# Signature    2    FALSE    FALSE    FALSE
# Signature    3    FALSE    FALSE    FALSE

5.2 Display

The S4 method of show prints the vital information, e.g., the number of signatures, and the number of positive and negative subjects, of an S4 'panel' object (Listing 1.5).

5.3 True & False Positives

The function sum1() and sum0() find the number of true- and false-positives per signature in a panel, respectively.

Listing 5.3 and Listing 5.4 find the number of true- and false-positives per signature in the panel Listing 1.5.

Listing 5.3: True positive per signature (Listing 1.5) (Zezulinski et al. 2025)

m |> 
  sum1() |>
  head(n = 3L)
# Signature    1 Signature    2 Signature    3 
#             12             10              6

Listing 5.4: False positive per signature (Listing 1.5) (Zezulinski et al. 2025)

m |> 
  sum0() |>
  head(n = 3L)
# Signature    1 Signature    2 Signature    3 
#              0              0              0

5.4 Subset

The S3 method subset.panel() 🚧

5.4.1 By Panel-Wise False Positives

The ordered panel Listing 2.4 controlled at one-or-less false positive per signature has a substantial false positive rate per ordered sub-panel (Table 2.2).

Listing 5.5 imposes an additional criterion on false positive per ordered sub-panel.

Listing 5.5: Example: Additional criteria on false positive per ordered sub-panel (Listing 2.4) (Zezulinski et al. 2025)

p1a = p1 |>
  subset(subset = cumsum0 <= 4L)
p1a
# Signature False(+) ≤1/31
# Panel of 13 Variant-Signatures from
# 50 positive subjects
# 31 negative subjects

Figure 5.1 (Listing 5.6) adds an additional exclusion box to the flow chart Figure 2.2.

Listing 5.6: Flow chart of Listing 5.5. (Zezulinski et al. 2025)

p1a |>
  plot()

Figure 5.1: Flow Chart of Listing 5.5. (Zezulinski et al. 2025)

Table 5.1 (Listing 5.7) drops the last few lines of Table 2.2 at the additional criteria on the false positive per ordered panel.

Listing 5.7: Variant-signatures in Listing 5.5. (Zezulinski et al. 2025)

p1a |> 
  flextable::as_flextable()

Table 5.1: Variant-Signatures in Ordered Panel Listing 5.5

Variant-Signature	Variant(s) in Signature	Individual Signature		Ordered Sub-Panel Signature False(+) ≤1/31
		True(+)	False(+)	Sub-Panel True(+)	Sub-Panel False(+)
Signature 1	chr17:7176997 C>T	12/50	0/31	24.0% =12/50	0.0% =0/31
Signature 340	chr18:51196776 G>C	12/50	1/31	40.0% =20/50	3.2% =1/31
Signature 2	chr11:118583356 G>A	10/50	0/31	56.0% =28/50	3.2% =1/31
Signature 341	chr5:141582348 C>G	8/50	1/31	66.0% =33/50	6.5% =2/31
Signature 342	chr17:43015806 C>T	7/50	1/31	72.0% =36/50	9.7% =3/31
Signature 343	chr19:1080497 C>A	7/50	1/31	74.0% =37/50	9.7% =3/31
Signature 5	chr3:9789723 C>T	5/50	0/31	78.0% =39/50	9.7% =3/31
Signature 346	chr19:38840293 C>T	5/50	1/31	80.0% =40/50	12.9% =4/31
Signature 7	chr1:29053204 TA>T	4/50	0/31	82.0% =41/50	12.9% =4/31
Signature 8	chr1:29053290 C>A chr1:29053291 A>G chr1:29053294 G>C chr1:29053295 C>T	4/50	0/31	84.0% =42/50	12.9% =4/31
Signature 11	chr17:7313221 G>A	4/50	0/31	86.0% =43/50	12.9% =4/31
Signature 14	chr19:10252358 G>C	4/50	0/31	88.0% =44/50	12.9% =4/31
Signature 17	chr19:55092500 C>T	4/50	0/31	90.0% =45/50	12.9% =4/31

Figure 5.2 (Listing 5.8) and Figure 5.3 (Listing 5.9) truncate the operating characteristic curve (Figure 3.1) and the pseudo receiver operating characteristic (ROC) curve (Figure 3.2) of the ordered panel Listing 2.4.

Listing 5.8: Figure: operating characteristics of Listing 2.1 and Listing 5.5. (Zezulinski et al. 2025)

panellist(p0, p1a) |>
  ggplot2::autoplot() +
  ggplot2::theme_minimal()

Figure 5.2: Operating Characteristics of Listing 2.1 and Listing 5.5. (Zezulinski et al. 2025)

Listing 5.9: Figure: Pseudo receiver operating characteristics of Listing 2.1 and Listing 5.5. (Zezulinski et al. 2025)

panellist(p0, p1a) |>
  ggplot2::autoplot(which = 'roc') +
  ggplot2::theme_minimal()

Figure 5.3: Pseudo ROC of Listing 2.1 and Listing 5.5. (Zezulinski et al. 2025)

5.5 Sort-By

The S3 method sort_by.panel() 🚧

5.6 Cumulative True & False Positives

The function cumsum1() finds the number of true positives, by a sub-panel of the first \(i\) signatures.

The function cumsum0() finds the number of false positives, by a sub-panel of the first \(i\) signatures.

The function cumsum1() and cumsum0() return a not-strictly increasing integer vector.

# example

5.7 Append Label

The function append_label() 🚧

5.8 Flow Chart

The S3 method plot.panel() creates a flow chart of the creation of a panel, using the consort package (Dayim 2024).

The S3 method plot.panel() is a wrapper of the function consort::consort_plot(), which returns an object of class 'consort'.

Examples using the S3 method plot.panel() are provided in Chapter 2.

Dayim, Alim. 2024. consort: Create Consort Diagram. https://CRAN.R-project.org/package=consort.

Zezulinski, Daniel, Maarouf A. Hoteit, David E. Kaplan, et al. 2025. “Detection of Circulating mRNA Variants in Hepatocellular Carcinoma Patients Using Targeted RNAseq.” Liver Cancer, ahead of print. https://doi.org/10.1159/000545366.