S2 (505(b)(2))

Diabetic Nephropathy

Library compounds whose LINCS consensus expression profiles most strongly anti-correlate one transcriptional signature associated with Diabetic Nephropathy, after a cytotoxicity-alignment filter and clinical-maturity ranking. Compounds our structural screen flags are labelled below rather than hidden. This is a property of that signature, not of the disease. Computational and hypothesis-generating, pending wet-lab. Method & limits →

Not tested. Twenty-three indications in our corpus hold two independently derived signatures. Comparing the twelve top candidates from each: 14 pairs shared no compound, 6 shared one, 3 shared two — indistinguishable from the 3,173 pairs built from different indications, where 60.1% likewise share nothing (p = 0.92, two-sided). But in every one of those pairs one signature was far too shallow to measure anything, so that outcome was expected before the test was run. It is not evidence that the method fails; it is evidence that we never ran a test capable of succeeding. Treat the names below as leads from one measurement, not as candidates for the disease.

Signature availability Diabetic Nephropathy

This indication is not ranked and not for sale. What follows is the measurement we hold and why it is not enough, published because it is the part of this work most likely to be useful to someone attempting the same thing.

Signature	CREEDS dz:223 (GSE1009, human)
Organism	human
Genes in the signature	599
Of those, inside the LINCS-978 landmark set	61
Position in our corpus of 118 signatures	27th percentile

The signature is not small — it is measured on a different gene space. It carries 599 genes, but our score is a rank correlation computed only over those it shares with the 978-gene LINCS landmark set, and that intersection is 61. At that depth, we have no measurement. At 100 shared genes 80.8% of draws reproduce the ranking; at 50 the result could not be separated from a signature whose gene labels had been shuffled. This signature sits between those two points.

The 61 landmark genes this signature covers

Direction as recorded in the source signature. This is the entire overlap the score was computed over, published so you can judge whether that subset is biologically coherent for this indication and re-run it yourself.

ALDOA ↑APP ↑BECN1 ↑CCND1 ↑CEBPA ↑CSNK1E ↑EIF4EBP1 ↑GNAI2 ↑HLA-DMA ↑HLA-DRA ↑HSPB1 ↑LGMN ↑PAX8 ↑PHGDH ↑S100A4 ↑SORBS3 ↑SPP1 ↑ARID5B ↓ATP1B1 ↓CEBPD ↓CIRBP ↓CLSTN1 ↓CSRP1 ↓DAG1 ↓DCTD ↓FAT1 ↓GADD45A ↓GTF2A2 ↓HSPA1A ↓HTRA1 ↓HYOU1 ↓ITGB5 ↓KLHL9 ↓KTN1 ↓LGALS8 ↓MIF ↓MLEC ↓MTHFD2 ↓MYL9 ↓MYLK ↓PGAM1 ↓PLSCR1 ↓PRCP ↓PSMB8 ↓PSME1 ↓RGS2 ↓RHOA ↓RPS6 ↓SCRN1 ↓SKP1 ↓SPAG7 ↓SPTAN1 ↓STAT1 ↓SYPL1 ↓TMED10 ↓TRAM2 ↓TSC22D3 ↓TXNRD1 ↓UBE2L6 ↓WFS1 ↓XBP1 ↓

Compounds this signature surfaced alphabetical · deliberately unranked

Shown so you can see what a 61-gene correlation produces, not so you can act on it. Ordering is withheld because at this depth the order is a tie-break artifact rather than a measurement. Our own structural screen and registered-trial counts are shown rather than hidden.

ceforanideSTRUCTURAL ALERT
cyclopentolate
glipizide
iopanoic acidSTRUCTURAL ALERT
lenalidomideSTRUCTURAL ALERT
meropenemSTRUCTURAL ALERT
procyclidine
thalidomideSTRUCTURAL ALERT

What would change this

A signature for this indication measured on a platform that overlaps the landmark set more deeply. Our floor is 300 genes; 200 is the deepest point at which we have actually measured reliability. We cannot tell you which genes would matter most here — that would need a curated disease-gene map for this indication, which we do not hold and are not going to assert. If you have a deeper signature for this disease and want it scored against our library, that is a conversation we are glad to have.

Ranking withheld for this indication

Our reversal score is a rank correlation computed over the genes this disease signature shares with the LINCS-978 landmark set. For Diabetic Nephropathy only 61 genes overlap, against a median of 916 across our corpus. A correlation over so few genes is not a reliable measurement, so we are not ranking or selling candidates for this indication until the signature is rebuilt against the landmark space. The candidate names above are shown for transparency and are not ordered by a trustworthy score.

How the score works & its limits →