Selection recovery versus purity in tissue profiling

Selection recovery versus purity in tissue profiling separates how many desired cells are retained from how strongly those cells dominate the collected material. In laser capture microdissection, high target-cell purity can coexist with low target recovery. Neither quantity alone describes the biochemical amount recovered or guarantees that the retained cell population represents every target state.

Let the starting specimen contain $N_T$ target and $N_B$ background cells, and suppose selection retains $n_T,n_B$. Definewhen the denominators are positive. $R_T$ measures target recovery and $P_T$ target purity. The analogous notions in a classified set are precision and recall, but here the quantities describe a physical collection process, not merely its digital labels.

For an original specimen containing 100 targets and 900 background cells, a broad selection recovering 60 targets and 40 background cells has $R_T=60\%$ and $P_T=60\%$. A stricter selection recovering 20 targets and one background cell instead has $R_T=20\%$ and $P_T\simeq 95.2\%$. It gives a cleaner mixture but only one third as many target cells as the first selection. The suitable balance depends on the downstream assay and the sampling question.

The enrichment can be expressed as an odds identity. If $p=N_T/(N_T+N_B)$ and the respective recovery fractions are $q_T,q_B>0$, thenThe ratio $q_T/q_B$ multiplies the starting target odds. This is a direct consequence of counting recovered targets and background, not a claim that neighbouring cells in a laser footprint are independently sampled.

Even perfect target purity can hide selection bias within the desired population. Suppose two target states initially occur equally often, but their recovery fractions are $0.9$ and $0.1$. Among recovered targets, the first state accounts for $0.9/(0.9+0.1)=90\%$. If its per-cell marker signal is 100 units and the second state's is 20, the selected mean is 92 units instead of the original 60, although neither state's own signal changed. Target identity, representative state sampling and gene expression are separate checks.

The original capture experiment demonstrated selected tissue transfer together with DNA/RNA amplification and enzyme recovery. Those biochemical tests establish useful recovered material, but do not supply every target/background count needed for the above recovery and purity measures.^[ref] Before/after images, target counts and matched assay-background controls provide complementary evidence. Target purity versus cell-state change in microdissection adds the distinction between changed composition and changed per-cell biology.