一、Spot-based / NGS-based
🔵 10x Visium FFPE / FF
解析度：55 µm spot（≈ 1–10 cells）
spot 間距：100 µm
覆蓋：全轉錄組（probe-based for FFPE）
面積：6.5 mm × 6.5 mm（標準）/ 11 mm × 11 mm (CytAssist)
定位：產業界用量最大、社群與工具支援最完整、適合初學者第一個 ST 實驗。
Resolution: 55 µm spot (≈ 1–10 cells)
Spot pitch: 100 µm
Coverage: whole transcriptome (probe-based for FFPE)
Area: 6.5 × 6.5 mm (standard) / 11 × 11 mm (CytAssist)
Niche: the industry default — largest community, most tools, ideal first ST experiment.
🔵 Visium HD FFPE / FF 2 µm
解析度：2 µm bin（連續無 gap）
常用 bin：2 / 8 / 16 µm，10x 推薦先用 8 µm 分析
覆蓋：~18 085 基因（probe-based）
定位：2024 上市、近單細胞解析度的 NGS-based ST，FFPE 相容、與 Seurat v5 / BANKSY 整合度高。
Resolution: 2 µm bins (continuous, no gap)
Common bins: 2 / 8 / 16 µm — 10x recommends starting at 8 µm
Coverage: ~18 085 genes (probe-based)
Niche: launched 2024 — near single-cell NGS-based ST, FFPE-compatible, tight Seurat v5 / BANKSY integration.
🔵 Slide-seq V2 FF
解析度：10 µm bead
覆蓋：全轉錄組
定位：學術/開源管道，beads packing 成連續面，但靈敏度比 Visium 低，需要更高定序深度。
Resolution: 10 µm beads
Coverage: whole transcriptome
Niche: academic / open pipeline; densely packed beads but lower sensitivity than Visium — needs deeper sequencing.
🔵 Stereo-seq (BGI) FF 500 nm
解析度：500 nm DNB（可下游 binning 到 25/50/100）
面積：支援 13 cm × 13 cm 超大 chip
覆蓋：全轉錄組
定位：大面積 + 高解析度，特別適合整顆器官切面或胚胎全胚層研究。
Resolution: 500 nm DNB spots (binned downstream to 25/50/100)
Area: up to 13 × 13 cm chips
Coverage: whole transcriptome
Niche: very large area + high resolution — ideal for whole-organ sections or whole-embryo studies.
二、Image-based / FISH-based
🟣 10x Xenium FFPE / FF subcellular
解析度：真正單細胞 / subcellular
Panel：Multi-Tissue (377)、5K Pan-Tissue (5 001)、自訂 panel
掃描範圍：12 mm × 24 mm
定位：FFPE 相容性極佳、社群快速成長、Xenium Analyzer 內建 segmentation。Panel 越大每基因偵測敏感度越低。
Resolution: true single-cell / subcellular
Panels: Multi-Tissue (377), 5K Pan-Tissue (5 001), custom
Scan area: 12 × 24 mm
Niche: excellent FFPE handling, fast-growing community, built-in segmentation. Larger panel ⇒ lower per-gene sensitivity.
🟣 Vizgen MERFISH / MERSCOPE FFPE / FF
解析度：subcellular
Panel：140–1 000 基因
掃描範圍：~1 cm²
定位：在高品質 RNA 上敏感度略勝 Xenium，學術廣泛應用，特別在腦圖譜（Allen Brain）上有大型參考資料。
Resolution: subcellular
Panel: 140–1 000 genes
Scan area: ~1 cm²
Niche: slightly higher sensitivity than Xenium on high-quality RNA; widely used in academia, especially the Allen Brain atlases.
🟣 NanoString CosMx SMI FFPE / FF + protein
解析度：subcellular
Panel：Whole Transcriptome (~18 000) / 6K / 1K 可選；可同時加 64 種 protein
定位：提供全轉錄組 image-based 選項，但每基因敏感度跟細胞通量需取捨；多模態 (RNA + protein) 是賣點。
Resolution: subcellular
Panel: Whole-transcriptome (~18 k) / 6K / 1K; up to 64 proteins simultaneously
Niche: only whole-transcriptome image-based option; trade-offs in per-gene sensitivity & throughput; multi-omic RNA + protein.
🟣 STARmap PLUS FF
解析度：單細胞
Panel：160–1 020 基因
定位：學術 in-situ sequencing 平台，特別在 3D 厚切片成像有優勢；自架平台、商業化程度低。
Resolution: single-cell
Panel: 160–1 020 genes
Niche: academic in-situ sequencing — strong on 3D thick-section imaging; in-house build, limited commercialization.
互動：在 trade-off 空間中找你需要的平台
X 軸 = 解析度 (µm，越小越精細)；Y 軸 = panel/transcriptome 大小（基因數，越大越廣）。圓圈大小 ∝ 掃描面積。
X-axis = resolution (µm, smaller = finer); Y-axis = panel size (genes, larger = broader). Bubble size ∝ scan area.
三、平台規格速查表
| 平台 | 家族 | 解析度 | 基因數 | FFPE | 典型用途 | ||
|---|---|---|---|---|---|---|---|
| Visium | Spot | 55 µm | ~18 k | ✓ | 入門首選；組織區域層級分析 | Beginner default; tissue-region analysis | |
| Visium HD | Spot | 2 / 8 µm | ~18 k | ✓ | 需要近單細胞解析的 NGS 路線 | When near-single-cell + NGS is needed | |
| Slide-seq V2 | Spot | 10 µm | ~30 k | ✗ | 學術自架管道；需高定序量 | Academic open pipeline; deep sequencing | |
| Stereo-seq | Spot | 500 nm | ~30 k | FF† | 超大樣本（整器官、整胚胎） | Very large samples (whole organ / embryo) | |
| Xenium | Image | subcellular | 377 / 5001 | ✓ | subcellular | FFPE 單細胞、要可重現的商業化方案 | FFPE single-cell, robust commercial workflow |
| MERSCOPE | Image | subcellular | 140–1 000 | ✓ | subcellular | 學術腦圖譜、高品質 RNA | Brain atlases, high-quality RNA |
| CosMx SMI | Image | subcellular | 1 k–18 k | ✓ | subcellular | image-based 全轉錄組 / RNA + protein 多模態 | Image-based whole-transcriptome / RNA + protein |
† Stereo-seq 對 FFPE 支援已有實驗版，但官方主流仍為 fresh frozen。
四、樣本品質的硬規則
- 新鮮冷凍：RIN ≥ 7 強烈建議。Visium fresh frozen 對 RNA 品質敏感。
- FFPE：看 DV200（> 200 nt 片段比例）。Visium FFPE 需 DV200 ≥ 30%；Xenium 較寬鬆。
- 定序深度：Visium FFPE 標準 25 k reads/spot 通常不夠，建議 100–120 k reads/spot 才能看到 transcript diversity。
- 切片厚度：Visium 5–10 µm；Xenium 5 µm；MERSCOPE 10 µm；STARmap 可達 100 µm 厚切。
- Fresh frozen: RIN ≥ 7 strongly recommended. Visium FF is RNA-quality-sensitive.
- FFPE: use DV200 (% fragments > 200 nt). Visium FFPE needs DV200 ≥ 30%; Xenium is more tolerant.
- Sequencing depth: the default 25 k reads/spot for Visium FFPE is rarely enough — aim for 100–120 k reads/spot to capture transcript diversity.
- Section thickness: Visium 5–10 µm; Xenium 5 µm; MERSCOPE 10 µm; STARmap up to 100 µm thick sections.
不同平台的讀檔範例
# Visium (含 H&E low-res) vis <- Load10X_Spatial("visium_out/") # Visium HD（同時載入多 bin） hd <- Load10X_Spatial("hd_out/binned_outputs/", bin.size = c(8, 16)) # Xenium（per-cell + per-transcript） xen <- LoadXenium("xenium_out/", fov = "fov") # MERSCOPE mer <- ReadVizgen("merfish_out/")
import scanpy as sc import squidpy as sq import spatialdata_io as sd_io # Visium adata = sc.read_visium("visium_out/") # Visium HD sdata_hd = sd_io.visium_hd("hd_out/") # Xenium / MERSCOPE / CosMx — 統一用 SpatialData sdata_xe = sd_io.xenium("xenium_out/") sdata_mer = sd_io.merscope("merscope_out/") sdata_cos = sd_io.cosmx("cosmx_out/")
📝 自我檢測
1. 你想做整個小鼠胚胎切片、看不同器官如何分化，最合適的平台？
1. You want to profile a whole mouse embryo section to study organ differentiation. Best platform?
2. 你已經知道要看的 60 個免疫標記基因，且樣本是 FFPE。最合適？
2. You have a defined 60-gene immune panel and FFPE tissue. Best fit?
3. 關於 FFPE 樣本的品質指標，下列敘述何者正確？
3. Which is correct about FFPE sample QC?