Compilation of AD-associated and non-AD genes AD-associated (positives) and non-AD (positives) genes are needed to build a machine learning model. First of all, we performed intensive hand-curation to identify confident AD-associated genes (positives) from various disease genes resources, including AlzGene1, AlzBase2, OMIM3, DisGenet4, DistiLD5, and UniProt6, Open Targets7, GWAS Catalog8, differentially expressed genes (DEGs) in ROSMAP9, and published literature. The curated genes from each resource as well as the corresponding criteria were provided in Table 1 in this file. As the AD-associated genes as well as their reliability vary across these resources, we used a vote strategy and selected only the genes that are present in at least two of the above-described resources to ensure higher reliability. In this way, we collected 147 AD-associated genes (Table 2). Of them, most (n=103, 70%) were associated with AD based on GWAS. Next, we selected a set of non-AD genes, (i.e. negatives), which refer to the genes that have no or minimal association with AD. The main idea of our method for non-AD gene selection was to remove any genes that exhibit potential associations with AD. Let Gp be the set of the 147 collected positives. We then selected a negative gene set GN containing genes that are potentially associated with AD in the following procedure. We first performed a Gene Ontology (GO) enrichment analysis for the positives using PANTHER10 and obtained 737 biological process terms (FDR<0.05). It is reasonable to assume that the genes annotated to these GO terms exhibit a potential association with AD compared with a random baseline because they were annotated to the same GO terms (functions) as the positive genes. In addition, we also collected genes that showed any potential association with AD from the above-described resources. These genes were then combined with the genes annotated to the GO terms, forming a set of genes of potential association with AD, denoted by GS. Next, the negative gene set was then calculated as GN = GA - GS, where GA represents all the genes in the network. By removing the 19471 genes identified in GS from GA (21122 genes), 1651 genes that were not associated with AD were identified and used as non-AD (negatives) genes. Table 1. Resources for compiling AD-associated genes. No. of Resources Genes Filtering Criteria genes PSEN2, HFE, NOS3, PLAU, CALHM1, A2M, PSEN1, OMIM3 All genes associated with AD 12 ADAM10, MPO, ABCA7, APOE, APP APOC1, BCL3, MS4A4A, FRMD4A, GAB2, MTHFD1L, CR1, EPHA1, CLU, PICALM, Genome-wide significant association DistiLD5 19 SORL1, CD33, ABCA7, with AD (p-value < 5.0×10-8) CEACAM16, GLIS3, CD2AP, SRRM4, BCHE, ZNF292 CLU, TNK1, APOE, SOAT1, PVRL2, TRAK2, CHRNB2, TFAM, CTSD, IL33, THRA, IL1RN, IL10, PICALM, STH, IL6, CD33, TFCP2, VLDLR, SORCS1, IL1B, GAB2, ABCA1, CTNNA3, APOC2, IL1A, BIN1, DAPK1, PON1, CST3, CCR2, The top ranked AD genes provided AlzGene11 GAPDHS, IL8, GRN, ECE1, 68 on the AlzGene website. OLR1, TF, LPL, LRP1, HHEX, GAPDH, LDLR, OTC, ARID5B, TNF, TOMM40, EXOC3L2, PGBD1, PRNP, BCAM, CALHM1, UBQLN1, PCK1, NEDD9, SORL1, ADAM10, MME, ENTPD7, FAS, CH25H, APOC1, ACE, CYP19A1, NCAPD2, MTHFR, APOC4, CHAT, CR1 APP, PSEN1, PSEN2, MAPT, APOE, TREM2, PLD3, SORL1, NOTCH3, CLU, PICALM, CR1, ABCA7, BIN1, CD33, CD2AP, MS4A6A, MS4A4E, NME8, Top ranked genes supported by AlzBase2 32 EPHA1, PTK2B, SLC24A4, genetic evidences CELF1, FERMT2, ZCWPW1, INPP5D, HLA-DRB1, HLA- DRB5, MEF2C, TP53INP1, CASS4, IGHV1-67 PSEN1, APP, PSEN2, APOE, Genes associated with AD and score DisGenet12 SORL1, MAPT, BDNF, IL1B, 12 > 0.3 BACE1, ACE, GSK3B, PLAU ADAM10, APOE, APP, MT-ND1, UniProt 6 MT-ND2, PSEN1, PSEN2, All genes associated with AD 8 SORL1 SLC30A6(ZNT6)13, 14, SLC30A4(ZNT4)13, 14, MAOB15, 16, CHRNA717, 18, IGF119, 20, VEGFA21, 22, ARC23, 24, PPARG25, 26, BCL227, 28, 29, IGF2R30, CASP331, 32, NECTIN2 (PVRL2)33, 34, 35, IGF230, 36, 37, VSNL138, 39, ACHE40, 41, 42, CYP46A143, 44, 45, IDE46, 47, 1. For GWAS studies: p- NPY48, 49, 50, IGF1R20, 51, value<5.0×10-8 Published PM20D152, APBB253, 54, ESR155, 2. For differential gene expression 56 literature 56, 57, PAXIP158, 59, CRH60, 61, studies on AD, the gene which is SOD262, 63, 64, SLC2A465, 66, differentially expressed in at least HMOX167, 68, DPYSL269, 70, two studies were included. INSR71, 72, DHCR2473, 74, 75, RELN76, 77, 78, 79, 80, BAX81, 82, 83, REST84, MEOX285, 86, EIF2AK387, 88, MAN2A189, 90, (ABI3, PLCG2)91, 92, DSG226, 93, MS4A294, 95,PPP1R3796, RELB33, 97, TREML298, (HESX1, ADAMTS4, CLNK, HS3ST1, CNTNAP2, ADAM10, APH1B, KAT8, ABI3, ALPK2, ECHDC3, SCIMP, SUZ12P1)99 p-value<5.0×10-8; from both GWAS- GWAS reported and GWAS-mapped genes 265 catalog https://www.ebi.ac.uk/gwas as provided in the GWAS catalog database. Open Genes with association score with AD 175 Targets7 https://www.opentargets.org/ > 0.5 ROSMAP See ref9 All DEGs with FDR<0.05 2614 Table 2. The list of 147 AD-associated genes. AD-associated genes APOE, SORL1, GAB2, CR1, PICALM, CLU, CD33, ABCA7, ADAM10, CD2AP, BIN1, APOC1, TOMM40, INPP5D, PSEN2, EPHA1, APP, MTHFD1L, CNTNAP2, HLA-DRB1, CASS4, BCAM, ABCA1, PTK2B, MS4A6A, FRMD4A, BCL3, SLC24A4, GLIS3, FERMT2, PSEN1, TREM2, ZCWPW1, EXOC3L2, MS4A4A, ACE, APOC4, BZW2, SUCLG2, APOB, SCIMP, SCARB1, RELB, CRY2, PVRL2, CLASRP, ADAMTS4, MMP3, UBE2L3, PPP1R37, ECHDC3, TCF7L2, IL6R, MS4A2, LIPG, MAN2A1, MAPT, ALDH1A2, ABI3, LILRA5, CELF1, PLCG2, HMGCR, OARD1, APH1B, APOC2, OR4S1, STAT4, MS4A4E, PVR, MT-ND2, HS3ST1, CCR2, VASP, CYP8B1, BLOC1S3, PPP1R13L, NFIC, NKPD1, INSR, CNTNAP5, BCAS3, BCHE, BCL2, NME8, CLPTM1, CLNK, UBQLN1, CLMN, IL1B, TRAPPC6A, VSNL1, SORCS1, PPARG, IGSF23, CRH, PSMA1, CHRNB2, FBXL7, CHRNA7, SPON1, MYO16, CHRNA2, VLDLR, KIR3DL2, KIT, HLA-DRB5, BACE1, HLA- DRA, DSG2, CALHM1, RBFOX1, HFE, PILRA, LRP4, HARBI1, TFCP2, CBLC, DPP10, SYNJ1, CDC25B, ACP2, ACHE, PACSIN3, MADD, ZNF652, GSK3B, PFDN1, RIN3, MARK4, GRIN2A, PDGFRB, MAPK8IP1, GRIN3B, CCRL2, ECE1, SCN1A, HBEGF, CACNA1G, CEACAM16, MMP13, ESR1, ALDH5A1, PLAU, SCN8A, CACNA2D1, MMP12 References 1. Bertram L, McQueen MB, Mullin K, Blacker D, Tanzi RE. Systematic meta-analyses of Alzheimer disease genetic association studies: the AlzGene database. Nat Genet 39, 17–23 (2007). 2. Bai Z, et al. AlzBase: an integrative database for gene dysregulation in Alzheimer’s disease. Mol Neurobiol 53, 310–319 (2016). 3. Hamosh A, Scott AF, Amberger JS, Bocchini CA, McKusick VA. Online Mendelian Inheritance in Man (OMIM), a knowledgebase of human genes and genetic disorders. Nucleic Acids Res 33, D514–D517 (2005). 4. Pinero J, et al. DisGeNET: a discovery platform for the dynamical exploration of human diseases and their genes. Database (Oxford) 2015, bav028 (2015). 5. Palleja A, Horn H, Eliasson S, Jensen LJ. DistiLD Database: diseases and traits in linkage disequilibrium blocks. Nucleic Acids Res 40, D1036–D1040 (2012). 6. Wu CH, et al. The Universal Protein Resource (UniProt): an expanding universe of protein information. Nucleic Acids Res 34, D187–D191 (2006). 7. Koscielny G, et al. Open Targets: a platform for therapeutic target identification and validation. Nucleic Acids Research 45, D985-D994 (2016). 8. MacArthur J, et al. The new NHGRI-EBI Catalog of published genome-wide association studies (GWAS Catalog). Nucleic Acids Research 45, D896-D901 (2016). 9. Canchi S, et al. Integrating Gene and Protein Expression Reveals Perturbed Functional Networks in Alzheimer’s Disease. Cell Reports 28, 1103-1116.e1104 (2019). 10. Mi H, Muruganujan A, Ebert D, Huang X, Thomas PD. PANTHER version 14: more genomes, a new PANTHER GO-slim and improvements in enrichment analysis tools. Nucleic Acids Res 47, D419–D426 (2018). 11. Alzforum. https://www.alzforum.org. 12. Piñero J, et al. DisGeNET: a comprehensive platform integrating information on human disease-associated genes and variants. Nucleic Acids Research 45, D833-D839 (2016). 13. Lyubartseva G, Smith JL, Markesbery WR, Lovell MA. Alterations of zinc transporter proteins ZnT-1, ZnT-4 and ZnT-6 in preclinical Alzheimer's disease brain. Brain pathology (Zurich, Switzerland) 20, 343-350 (2010). 14. Smith JL, Xiong S, Markesbery WR, Lovell MA. Altered expression of zinc transporters-4 and -6 in mild cognitive impairment, early and late Alzheimer’s disease brain. Neuroscience 140, 879-888 (2006). 15. Schedin-Weiss S, et al. Monoamine oxidase B is elevated in Alzheimer disease neurons, is associated with γ-secretase and regulates neuronal amyloid β-peptide levels. Alzheimer's Research & Therapy 9, 57 (2017). 16. Riederer P, Danielczyk W, Grünblatt E. Monoamine Oxidase-B Inhibition in Alzheimer’s Disease. NeuroToxicology 25, 271-277 (2004). 17. Clarelli F, et al. CHRNA7 Gene and Response to Cholinesterase Inhibitors in an Italian Cohort of Alzheimer's Disease Patients. J Alzheimers Dis 52, 1203-1208 (2016). 18. Carson R, et al. Alpha7 nicotinic acetylcholine receptor gene and reduced risk of Alzheimer's disease. J Med Genet 45, 244-248 (2008). 19. Westwood AJ, et al. Insulin-like growth factor-1 and risk of Alzheimer dementia and brain atrophy. Neurology 82, 1613-1619 (2014). 20. Ostrowski PP, Barszczyk A, Forstenpointner J, Zheng W, Feng Z-P. Meta-Analysis of Serum Insulin-Like Growth Factor 1 in Alzheimer’s Disease. PLOS ONE 11, e0155733 (2016). 21. Chiappelli M, et al. VEGF gene and phenotype relation with Alzheimer's disease and mild cognitive impairment. Rejuvenation research 9, 485-493 (2006). 22. Del Bo R, Ghezzi S, Scarpini E, Bresolin N, Comi GP. VEGF genetic variability is associated with increased risk of developing Alzheimer's disease. Journal of the neurological sciences 283, 66-68 (2009). 23. Bi R, et al. The Arc Gene Confers Genetic Susceptibility to Alzheimer's Disease in Han Chinese. Mol Neurobiol 55, 1217-1226 (2018). 24. Ridge PG, Mukherjee, S., Crane, P. K., Kauwe, J. S. & Alzheimer's Disease Genetics Consortium. Alzheimer's disease: analyzing the missing heritability.