Checkpoints and Grading Rubric

The assignment is worth 100 points: Protein 25, single-cell 25, spatial 25, Agent Audit 15, and reproducibility/report 10. Use the checkpoint IDs below to track your progress. In report.pdf, add a short evidence index such as P2 → Figure 2, Table 1, protein/measure.py.

How points are awarded: Full credit requires clear results and enough code or source information to check them. Incomplete but valid evidence earns partial credit; missing or unverifiable evidence earns 0. There is no required biological answer or target statistic. A well-supported negative or uncertain result can earn full credit.

Task 1: Protein mutation analysis — 25 points

Checkpoint	Show in your submission	Points
P1 · Map the variants	Identify the protein, verify each variant against its sequence, and map all candidate residues to the chosen structure/model.	5
P2 · Measure structural evidence	Define and report reproducible measurements for each candidate, or explain and document a justified alternative where coordinates are unavailable.	7
P3 · Interpret with evidence	Compare candidate mechanisms using observed results and cited sources; separate observations from predictions and state uncertainty.	8
P4 · Communicate the comparison	Provide an annotated structure of the assigned protein and one table covering every candidate.	5

Task 2: Single-cell analysis — 25 points

Checkpoint	Show in your submission	Points
S1 · Baseline and QC	Report QC decisions, analysis settings and seed, cluster sizes, and a cluster-labeled UMAP.	6
S2 · Cell-type evidence	Use markers computed from your PBMC3k data and cited biology to support major labels; do not use predefined labels.	7
S3 · Resolution check	Compare at least three Leiden resolutions with other settings fixed; quantify splits/merges and discuss label stability.	5
S4 · Downsampling check	Rerun a fixed-seed 60–80% cell sample; compare retained cells quantitatively and discuss label stability.	4
S5 · Ambiguous population	Investigate one unclear or unstable population with marker comparisons and a reason for the uncertainty.	3

Task 3: Spatial transcriptomics — 25 points

Checkpoint	Show in your submission	Points
T1 · Expression-only domains	Cluster using expression without spatial information; show the same labels in expression space and on tissue, with settings and spot counts.	5
T2 · Spatial organization	Define the physical-neighbor graph and report an interpreted numerical spatial statistic.	5
T3 · Spatial genes	State the tested gene set and method, rank spatially variable genes, and map at least two genes on tissue.	4
T4 · Compare gene rankings	Compare differential-expression and spatial rankings for the same genes, with examples that explain their differences.	4
T5 · Negative control	Use at least 100 fixed-seed randomizations; compare the same observed and null statistic, report an empirical comparison, and note a limitation.	7

Agent Audit — 15 points

For each task (5 points each), show: sources used (1), what the agent computed versus retrieved (1), one independent check and its result (2), and one weakness or limitation you evaluated (1).

Reproducibility and report — 10 points

Two points each for: ordered code/notebooks with a run README; data sources and software versions; settings and random seeds; labeled figures/tables tied to conclusions; and a readable report.pdf with all three Agent Audits and the checkpoint evidence index.

Before submitting, mark each checkpoint complete or needs revision and give the page or file containing its evidence.