BoltzMaker Report | Marc C. Deller

Campaign summary

Field	Value	Details
Input file	boltz_input.md	boltz_input.md
Proteins	1	EGFR (327 aa)
Co-folded partners	0	none
Ligands	1	FRAG1 (SMILES)
Pockets	0	none -- every ligand folded without a site constraint
Apo structure references	0	none -- no apo-vs-holo comparison in this campaign
Predictions	1	EGFR_FRAG1
Predict affinity	yes	pIC50 predicted for every target
Ligand chemistry	clean	no stereo/protonation/fragment concerns detected
Boltz predict runtime	5m 23s	single run invocation
Accelerator	gpu	gpu = Metal/CUDA backend used; cpu = no GPU available
Workers	2	parallel data-loading workers (Boltz's own default is 2)
MPS watermark	1.0	PYTORCH_MPS_HIGH_WATERMARK_RATIO cap -- lower avoids swap on Apple unified memory
Max parallel samples	1	Boltz's own --max_parallel_samples

Summary table

Identity			Affinity		Confidence					Interactions	Structure
Protein	Ligand	Summary	Binder p	pIC50	Score	pTM	ipTM	Lig ipTM	pLDDT	Phobic	CIF
EGFR	FRAG1		0.64	6.21 ± 0.02	0.92	0.94	0.90	0.90	0.93	1	CIF

Download full CSV · Download summary CSV

🎯 affinity · 🛡️ confidence: Likely binder (p ≥ 0.7) Uncertain (0.3-0.7) Likely non-binder (p < 0.3) High confidence (≥ 0.7) Moderate confidence (0.5-0.7) Low confidence (< 0.5)

Pockets

No named pockets in this campaign -- every ligand was placed without a site constraint.

Pocket	Protein	Ligands	Targets	Contacts
Unconstrained	EGFR	FRAG1	1	none -- ligand placed freely

Ligand preparation

No stereocentre, protonation-state, or disconnected-fragment concerns detected.

Ligand structures

Sundefined stereocentreAcarboxylic acidNprimary/secondary aminePhphenolSO3sulfonic acidsaltsalt/disconnected fragment

FRAG1

MW 147 · cLogP 1.8 · TPSA 29

Download PDF · Download SMILES

Scaffolds: Bemis-Murcko, exact match first, then Tanimoto-clustered (Morgan r=2, 2048-bit, threshold 0.60) whole-group MCS as a verified fallback. Minimum highlighted substructure size: 8 heavy atoms. Stereocentre/ionizable-group highlighting from this campaign's own ligand-preparation check (see above).

Ranked predicted pIC50

Ranked confidence

pIC50 vs confidence score

Interaction counts by type

pIC50 vs binder probability

EGFR: residue interaction fingerprint

EGFR_FRAG1: binding site

Download PyMOL session

Download image

Interaction	Residue	Number	Chain	Distance
hydrophobic	PHE	28	A	3.99

Download CSV

Secondary structure shifts (apo vs holo)

Family	Status	Detail
EGFR	OK	7 motif row(s) across 1 target(s), annotator=kinase

1 target(s), 7 motif(s) compared
Mean Ca RMSD: 0.27 A (median 0.23 A) — largest shift: 0.53 A at EGFR_FRAG1 / alphaC_Glu
Mean centroid shift: 0.19 A
Flagged phi/psi outlier residues: 8
Kinase state changes detected: 0 DFG, 0 alphaC

Identity						Shift			Backbone	Kinase state
Family	Target	Ligand	Motif	Kind	Source	N res	RMSD (A)	Centroid delta (A)	Flagged phi/psi	DFG apo	DFG holo	DFG delta	alphaC apo	alphaC holo	alphaC delta
EGFR	EGFR_FRAG1	FRAG1	DFG	loop	kinase	3	0.21	0.16	0	out	out	False	out	out	False
EGFR	EGFR_FRAG1	FRAG1	catalytic_loop	loop	kinase	3	0.26	0.24	0	out	out	False	out	out	False
EGFR	EGFR_FRAG1	FRAG1	hinge	loop	kinase	3	0.16	0.13	0	out	out	False	out	out	False
EGFR	EGFR_FRAG1	FRAG1	pocket_scaffold	loop	kinase	73	0.47	0.00	8	out	out	False	out	out	False
EGFR	EGFR_FRAG1	FRAG1	alphaC_Glu	point	kinase	1	0.53	0.53	0	out	out	False	out	out	False
EGFR	EGFR_FRAG1	FRAG1	catalytic_Lys	point	kinase	1	0.06	0.06	0	out	out	False	out	out	False
EGFR	EGFR_FRAG1	FRAG1	gatekeeper	point	kinase	1	0.23	0.23	0	out	out	False	out	out	False

Download CSV

Per-motif Ca RMSD

Motif x target RMSD

Landlord narration

1 of 1 target summaries written on-device by the Apple Neural Engine; 0 rendered from the template. Every number here is checked against the figures the analysis computed; a summary stating one it was not given was replaced by the template.

This campaign tested the ability of the EGFR_FRAG1 target to predict the experimental pic50 of the ligand FRAG1. The results indicate that the target is well determined and can be trusted, as indicated by the well determined confidence, and the predicted pic50 of 6.21 is the best match for the experimental pic50 of 0.02.

Targets	1 predicted, 1 with a ligand
Receptors	EGFR
Ligands	FRAG1
Verdicts	1 proceed
Confidence	1 well determined
Flagged	0 of 1
Pose validated	no experimental structure to compare against

Key findings

Of 1 targets, 1 are marked proceed, 0 caution and 0 discard.
1 of 1 targets are well determined; 0 are not.

Highest predicted potency

FRAG1 on EGFR_FRAG1, predicted pIC50 6.21

Caveats

FRAG1 was omitted from the target structure because it shows no contacts with the receptor, and its predicted potency is higher than the experimental pic50 of 0.02. There are no other flags in the JSON, and nothing else is amiss. Therefore, nothing else can be said about the target structure.

Per target

Target	Verdict	Summary	Caveat	Written by
`EGFR_FRAG1`	proceed	This structure is well determined, and therefore should be trusted. FRAG1 shows no contacts with the receptor, and its predicted potency is 6.21, which is higher than the experimental pic50 of 0.02. The predicted potency is the best match for the experimental pic50, and therefore the predicted structure is very likely to reproduce the experimental structure. There are no other ligands supplied in the JSON. FRAG1 was therefore omitted from the target structure.	FRAG1 was omitted from the target structure because it shows no contacts with the receptor, and its predicted potency is higher than the experimental pic50 of 0.02. There are no other flags in the JSON, and nothing else is amiss. Therefore, nothing else can be said about the target structure.	model