Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Mar 1.
Published in final edited form as: Am J Med Genet A. 2022 Dec 9;191(3):659–671. doi: 10.1002/ajmg.a.63060

Perspectives on the future of dysmorphology

Benjamin D Solomon 1, Margaret P Adam 2, Chin-To Fong 3, Katta M Girisha 4, Judith G Hall 5,6, Anna CE Hurst 7, Peter M Krawitz 8, Shahida Moosa 9,10, Shubha R Phadke 11, Cedrik Tekendo-Ngongang 1, Tara L Wenger 12
PMCID: PMC9928773  NIHMSID: NIHMS1851434  PMID: 36484420

Acknowledgements:

BDS is the co-Editor-in-Chief and KMG, ACEH, and SM are and TLW was previously Associate Editor(s) of the American Journal of Medical Genetics. BDS, MA, ACEH, and PK have collaborated or served roles with FDNA. This work was supported in part by the Intramural Research Program of the National Human Genome Research Institute, National Institutes of Health.

Abbreviations

AI

artificial intelligence

ES

exome sequencing

GS

genome sequencing

gnomAD

The Genome Aggregation Database

LVOD

Leiden Open Variation Database

NHGRI

National Human Genome Research Institute

NICU

neonatal intensive care unit

VACTERL

vertebral anomalies, anal atresia, cardiac anomalies, tracheo-esophageal fistula, renal anomalies, limb anomalies

VUS

variant of uncertain significance

References

  1. Abacan M, Alsubaie L, Barlow-Stewart K, Caanen B, Cordier C, Courtney E, … Wicklund C (2019). The Global State of the Genetic Counseling Profession. Eur J Hum Genet, 27(2), 183–197. 10.1038/s41431-018-0252-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Adam AP, Curry CJ, Hall JG, Keppler-Noreuil KM, Adam MP, & Dobyns WB (2020). Recurrent constellations of embryonic malformations re-conceptualized as an overlapping group of disorders with shared pathogenesis. Am J Med Genet A, 182(11), 2646–2661. 10.1002/ajmg.a.61847 [DOI] [PubMed] [Google Scholar]
  3. Adam M, & Hudgins L (2003). The importance of minor anomalies in the evaluation of the newborn. NeoReviews, 4(4), e99–e104. [Google Scholar]
  4. Adam MP, Hudgins L, Carey JC, Hall BD, Coleman K, Gripp KW, … Graham JM Jr. (2009). Preaxial hallucal polydactyly as a marker for diabetic embryopathy. Birth Defects Res A Clin Mol Teratol, 85(1), 13–19. 10.1002/bdra.20503 [DOI] [PubMed] [Google Scholar]
  5. Allanson JE, Biesecker LG, Carey JC, & Hennekam RC (2009). Elements of morphology: introduction. Am J Med Genet A, 149a(1), 2–5. 10.1002/ajmg.a.32601 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Baldridge D, Spillmann RC, Wegner DJ, Wambach JA, White FV, Sisco K, … Grange DK (2020). Phenotypic expansion of KMT2D-related disorder: Beyond Kabuki syndrome. Am J Med Genet A, 182(5), 1053–1065. 10.1002/ajmg.a.61518 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bamshad MJ, Nickerson DA, & Chong JX (2019). Mendelian Gene Discovery: Fast and Furious with No End in Sight. Am J Hum Genet, 105(3), 448–455. 10.1016/j.ajhg.2019.07.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bastarache L, Hughey JJ, Hebbring S, Marlo J, Zhao W, Ho WT, … Denny JC (2018). Phenotype risk scores identify patients with unrecognized Mendelian disease patterns. Science, 359(6381), 1233–1239. 10.1126/science.aal4043 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Biesecker LG, Aase JM, Clericuzio C, Gurrieri F, Temple IK, & Toriello H (2009). Elements of morphology: standard terminology for the hands and feet. Am J Med Genet A, 149A(1), 93–127. 10.1002/ajmg.a.32596 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Biesecker LG, Adam MP, Alkuraya FS, Amemiya AR, Bamshad MJ, Beck AE, … Zarate YA (2021). A dyadic approach to the delineation of diagnostic entities in clinical genomics. Am J Hum Genet, 108(1), 8–15. 10.1016/j.ajhg.2020.11.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Biesecker LG, Green ED, Manolio T, Solomon BD, & Curtis D (2021). Should all babies have their genome sequenced at birth? BMJ, 375, n2679. 10.1136/bmj.n2679 [DOI] [PubMed] [Google Scholar]
  12. Biesecker LG, & Green RC (2014). Diagnostic clinical genome and exome sequencing. N Engl J Med, 370(25), 2418–2425. 10.1056/NEJMra1312543 [DOI] [PubMed] [Google Scholar]
  13. Campbell IM, Crowley TB, Keena B, Donoghue S, McManus ML, & Zackai EH (2022). The experience of one pediatric geneticist with telemedicine-based clinical diagnosis. Am J Med Genet A. 10.1002/ajmg.a.62920 [DOI] [PubMed] [Google Scholar]
  14. Cavalcanti DP, Fano V, Mellado C, Lacarrubba-Flores MDJ, Silveira C, Silveira KC, … Ferreira CR (2020). Skeletal dysplasias in Latin America. Am J Med Genet C Semin Med Genet, 184(4), 986–995. 10.1002/ajmg.c.31861 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Cuvertino S, Hartill V, Colyer A, Garner T, Nair N, Al-Gazali L, … Banka S (2020). A restricted spectrum of missense KMT2D variants cause a multiple malformations disorder distinct from Kabuki syndrome. Genet Med, 22(5), 867–877. 10.1038/s41436-019-0743-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Duong D, Hu P, Tekendo-Ngongang C, Hanchard SEL, Liu S, Solomon BD, & Waikel RL (2022). Neural Networks for Classification and Image Generation of Aging in Genetic Syndromes. Front Genet, 13, 864092. 10.3389/fgene.2022.864092 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Ekure EN, Adeyemo A, Liu H, Sokunbi O, Kalu N, Martinez AF, … Kruszka P (2021). Exome Sequencing and Congenital Heart Disease in Sub-Saharan Africa. Circ Genom Precis Med, 14(1), e003108. 10.1161/CIRCGEN.120.003108 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Fattahi Z, Beheshtian M, Mohseni M, Poustchi H, Sellars E, Nezhadi SH, … Najmabadi H (2019). Iranome: A catalog of genomic variations in the Iranian population. Hum Mutat, 40(11), 1968–1984. 10.1002/humu.23880 [DOI] [PubMed] [Google Scholar]
  19. Fokkema IF, Taschner PE, Schaafsma GC, Celli J, Laros JF, & den Dunnen JT (2011). LOVD v.2.0: the next generation in gene variant databases. Hum Mutat, 32(5), 557–563. 10.1002/humu.21438 [DOI] [PubMed] [Google Scholar]
  20. The GenomeAsia 100K Project enables genetic discoveries across Asia. (2019). Nature, 576(7785), 106–111. 10.1038/s41586-019-1793-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gordon CT, Xue S, Yigit G, Filali H, Chen K, Rosin N, … Reversade B (2017). De novo mutations in SMCHD1 cause Bosma arhinia microphthalmia syndrome and abrogate nasal development. Nat Genet, 49(2), 249–255. 10.1038/ng.3765 [DOI] [PubMed] [Google Scholar]
  22. Green ED, Gunter C, Biesecker LG, Di Francesco V, Easter CL, Feingold EA, … Manolio TA (2020). Strategic vision for improving human health at The Forefront of Genomics. Nature, 586(7831), 683–692. 10.1038/s41586-020-2817-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Gripp KW, Slavotinek AM, Hall JG, & Allanson JE (2013). Handbook of physical measurements.
  24. Grody WW (2019). The transformation of medical genetics by clinical genomics: hubris meets humility. Genet Med, 21(9), 1916–1926. 10.1038/s41436-019-0450-0 [DOI] [PubMed] [Google Scholar]
  25. Gupta R, Gupta N, Nampoothiri S, Mandal K, Kishore Y, Sharma P, … Phadke SR (2016). Smith-Magenis Syndrome: Face Speaks. Indian J Pediatr, 83(6), 589–593. 10.1007/s12098-015-1940-y [DOI] [PubMed] [Google Scholar]
  26. Gurovich Y, Hanani Y, Bar O, Nadav G, Fleischer N, Gelbman D, … Gripp KW (2019). Identifying facial phenotypes of genetic disorders using deep learning. Nat Med, 25(1), 60–64. 10.1038/s41591-018-0279-0 [DOI] [PubMed] [Google Scholar]
  27. Hall J, Allanson J, Gripp K, & Slavotinek A (2007). Handbook of Physical Measurements Second Edition: Oxford university press. [Google Scholar]
  28. Hall JG (1985). The study of heterogeneity and natural history--the importance of this type of research with chondrodystrophies as an example. Prog Clin Biol Res, 177, 321–334. https://www.ncbi.nlm.nih.gov/pubmed/4011606 [PubMed] [Google Scholar]
  29. Hall JG (2003). A clinician’s plea. Nat Genet, 33(4), 440–442. 10.1038/ng0403-440 [DOI] [PubMed] [Google Scholar]
  30. Hallowell N, Parker M, & Nellaker C (2019). Big data phenotyping in rare diseases: some ethical issues. Genet Med, 21(2), 272–274. 10.1038/s41436-018-0067-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Harms FL, Girisha KM, Hardigan AA, Kortüm F, Shukla A, Alawi M, … Kutsche K (2017). Mutations in EBF3 Disturb Transcriptional Profiles and Cause Intellectual Disability, Ataxia, and Facial Dysmorphism. Am J Hum Genet, 100(1), 117–127. 10.1016/j.ajhg.2016.11.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Hartley T, Lemire G, Kernohan KD, Howley HE, Adams DR, & Boycott KM (2020). New Diagnostic Approaches for Undiagnosed Rare Genetic Diseases. Annu Rev Genomics Hum Genet, 21, 351–372. 10.1146/annurev-genom-083118-015345 [DOI] [PubMed] [Google Scholar]
  33. Hennekam RC, & Biesecker LG (2012). Next-generation sequencing demands next-generation phenotyping. Hum Mutat, 33(5), 884–886. 10.1002/humu.22048 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Hennekam RC, Biesecker LG, Allanson JE, Hall JG, Opitz JM, Temple IK, … Elements of Morphology, C. (2013). Elements of morphology: general terms for congenital anomalies. Am J Med Genet A, 161A(11), 2726–2733. 10.1002/ajmg.a.36249 [DOI] [PubMed] [Google Scholar]
  35. Hsieh TC, Bar-Haim A, Moosa S, Ehmke N, Gripp KW, Pantel JT, … Krawitz PM (2022). GestaltMatcher facilitates rare disease matching using facial phenotype descriptors. Nat Genet, 54(3), 349–357. 10.1038/s41588-021-01010-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Hsieh TC, Mensah MA, Pantel JT, Aguilar D, Bar O, Bayat A, … Krawitz PM (2019). PEDIA: prioritization of exome data by image analysis. Genet Med, 21(12), 2807–2814. 10.1038/s41436-019-0566-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Hurst ACE (2018). Facial recognition software in clinical dysmorphology. Curr Opin Pediatr, 30(6), 701–706. 10.1097/mop.0000000000000677 [DOI] [PubMed] [Google Scholar]
  38. Hurst ACE, & Robin NH (2020). Dysmorphology in the Era of Genomic Diagnosis. J Pers Med, 10(1). 10.3390/jpm10010018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Jaganathan K, Kyriazopoulou Panagiotopoulou S, McRae JF, Darbandi SF, Knowles D, Li YI, … Farh KK (2019). Predicting Splicing from Primary Sequence with Deep Learning. Cell, 176(3), 535–548 e524. 10.1016/j.cell.2018.12.015 [DOI] [PubMed] [Google Scholar]
  40. Jenkins BD, Fischer CG, Polito CA, Maiese DR, Keehn AS, Lyon M, … Watson MS (2021). The 2019 US medical genetics workforce: a focus on clinical genetics. Genet Med. 10.1038/s41436-021-01162-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Johnston JJ, Lewis KL, Ng D, Singh LN, Wynter J, Brewer C, … Biesecker LG (2015). Individualized iterative phenotyping for genome-wide analysis of loss-of-function mutations. Am J Hum Genet, 96(6), 913–925. 10.1016/j.ajhg.2015.04.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Kandasamy S, Saxena D, Kishore Y, & Phadke SR (2014). Williams syndrome: a case series. Indian Pediatr, 51(5), 411–412. 10.1007/s13312-014-0416-x [DOI] [PubMed] [Google Scholar]
  43. Karczewski KJ, Francioli LC, Tiao G, Cummings BB, Alföldi J, Wang Q, … MacArthur DG (2020). The mutational constraint spectrum quantified from variation in 141,456 humans. Nature, 581(7809), 434–443. 10.1038/s41586-020-2308-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Kausthubham N, Shukla A, Gupta N, Bhavani GS, Kulshrestha S, Das Bhowmik A, … Girisha KM (2021). A data set of variants derived from 1455 clinical and research exomes is efficient in variant prioritization for early-onset monogenic disorders in Indians. Hum Mutat, 42(4), e15–e61. 10.1002/humu.24172 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Kingsmore SF, Cakici JA, Clark MM, Gaughran M, Feddock M, Batalov S, … Investigators, R. (2019). A Randomized, Controlled Trial of the Analytic and Diagnostic Performance of Singleton and Trio, Rapid Genome and Exome Sequencing in Ill Infants. Am J Hum Genet, 105(4), 719–733. 10.1016/j.ajhg.2019.08.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Kruszka P, Tekendo-Ngongang C, & Muenke M (2019). Diversity and dysmorphology. Current opinion in pediatrics, 31(6), 702–707. [DOI] [PubMed] [Google Scholar]
  47. Lander ES, Linton LM, Birren B, Nusbaum C, Zody MC, Baldwin J, … International Human Genome Sequencing, C. (2001). Initial sequencing and analysis of the human genome. Nature, 409(6822), 860–921. 10.1038/35057062 [DOI] [PubMed] [Google Scholar]
  48. Landrum MJ, Lee JM, Benson M, Brown GR, Chao C, Chitipiralla S, … Maglott DR (2018). ClinVar: improving access to variant interpretations and supporting evidence. Nucleic Acids Res, 46(D1), D1062–d1067. 10.1093/nar/gkx1153 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Miller BC, Goldenberg AJ, & Bonhomme NF (2021). Digital peer-to-peer information seeking and sharing: Opportunities for education and collaboration in newborn screening. Am J Med Genet C Semin Med Genet, 187(1), 64–69. 10.1002/ajmg.c.31884 [DOI] [PubMed] [Google Scholar]
  50. Mistri M, Mehta S, Solanki D, Kamate M, Gupta N, Kabra M, … Sheth J (2019). Identification of novel variants in a large cohort of children with Tay-Sachs disease: An initiative of a multicentric task force on lysosomal storage disorders by Government of India. J Hum Genet, 64(10), 985–994. 10.1038/s10038-019-0647-8 [DOI] [PubMed] [Google Scholar]
  51. Naqvi S, Sleyp Y, Hoskens H, Indencleef K, Spence JP, Bruffaerts R, … Claes P (2021). Shared heritability of human face and brain shape. Nat Genet, 53(6), 830–839. 10.1038/s41588-021-00827-w [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Niikawa N, Kuroki Y, Kajii T, Matsuura N, Ishikiriyama S, Tonoki H, … et al. (1988). Kabuki make-up (Niikawa-Kuroki) syndrome: a study of 62 patients. Am J Med Genet, 31(3), 565–589. 10.1002/ajmg.1320310312 [DOI] [PubMed] [Google Scholar]
  53. Poplin R, Chang PC, Alexander D, Schwartz S, Colthurst T, Ku A, … DePristo MA (2018). A universal SNP and small-indel variant caller using deep neural networks. Nat Biotechnol, 36(10), 983–987. 10.1038/nbt.4235 [DOI] [PubMed] [Google Scholar]
  54. Robinson PN, Köhler S, Bauer S, Seelow D, Horn D, & Mundlos S (2008). The Human Phenotype Ontology: a tool for annotating and analyzing human hereditary disease. Am J Hum Genet, 83(5), 610–615. 10.1016/j.ajhg.2008.09.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Shalev SA, & Hall JG (2004). Behavioral pattern profile: a tool for the description of behavior to be used in the genetics clinic. Am J Med Genet A, 128A(4), 389–395. 10.1002/ajmg.a.30192 [DOI] [PubMed] [Google Scholar]
  56. Shur N, Atabaki SM, Kisling MS, Tabarani A, Williams C, Fraser JL, … Rare Disease, I. (2021). Rapid deployment of a telemedicine care model for genetics and metabolism during COVID-19. Am J Med Genet A, 185(1), 68–72. 10.1002/ajmg.a.61911 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Slavotinek AM, & Solomon BD (2020). Going forward in a new world. Am J Med Genet A, 182(7), 1553–1554. 10.1002/ajmg.a.61715 [DOI] [PubMed] [Google Scholar]
  58. Smith DW (1966). Dysmorphology (teratology). J Pediatr, 69(6), 1150–1169. 10.1016/s0022-3476(66)80311-6 [DOI] [PubMed] [Google Scholar]
  59. Snir M, Nazareth S, Simmons E, Hayward L, Ashcraft K, Bristow SL, … Aradhya S (2021). Democratizing genomics: Leveraging software to make genetics an integral part of routine care. Am J Med Genet C Semin Med Genet, 187(1), 14–27. 10.1002/ajmg.c.31866 [DOI] [PubMed] [Google Scholar]
  60. Solomon BD (2021a). Can artificial intelligence save medical genetics? Am J Med Genet A. 10.1002/ajmg.a.62538 [DOI] [PubMed] [Google Scholar]
  61. Solomon BD (2021b). Evolving technologies in medical genetics and genomics. Am J Med Genet C Semin Med Genet, 187(1), 5–6. 10.1002/ajmg.c.31891 [DOI] [PubMed] [Google Scholar]
  62. Tekendo-Ngongang C, Dahoun S, Nguefack S, Gimelli S, Sloan-Bena F, & Wonkam A (2014). Challenges in clinical diagnosis of williams-beuren syndrome in sub-saharan africans: case reports from cameroon. Mol Syndromol, 5(6), 287–292. 10.1159/000369421 [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Tekendo-Ngongang C, Owosela B, Fleischer N, Addissie YA, Malonga B, Badoe E, … Kruszka P (2020). Rubinstein-Taybi syndrome in diverse populations. Am J Med Genet A, 182(12), 2939–2950. 10.1002/ajmg.a.61888 [DOI] [PubMed] [Google Scholar]
  64. Tise CG (2021). Medical genetics training in the COVID-19 era: A resident’s perspective. Am J Med Genet A, 185(10), 2861–2862. 10.1002/ajmg.a.62379 [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Venter JC, Adams MD, Myers EW, Li PW, Mural RJ, Sutton GG, … Zhu X (2001). The sequence of the human genome. Science, 291(5507), 1304–1351. 10.1126/science.1058040 [DOI] [PubMed] [Google Scholar]
  66. Wenger TL, Bly RA, Wu N, Albert CM, Park J, Shieh J, … Dobyns WB (2020). Activating variants in PDGFRB result in a spectrum of disorders responsive to imatinib monotherapy. Am J Med Genet A, 182(7), 1576–1591. 10.1002/ajmg.a.61615 [DOI] [PubMed] [Google Scholar]
  67. Wonkam A (2021). Sequence three million genomes across Africa. Nature, 590(7845), 209–211. 10.1038/d41586-021-00313-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Zhang M, Wu S, Du S, Qian W, Chen J, Qiao L, … Wang S (2022). Genetic variants underlying differences in facial morphology in East Asian and European populations. Nat Genet, 54(4), 403–411. 10.1038/s41588-022-01038-7 [DOI] [PubMed] [Google Scholar]
  69. Zhang P, Luo H, Li Y, Wang Y, Wang J, Zheng Y, … He S (2021). NyuWa Genome resource: A deep whole-genome sequencing-based variation profile and reference panel for the Chinese population. Cell Rep, 37(7), 110017. 10.1016/j.celrep.2021.110017 [DOI] [PubMed] [Google Scholar]
Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Benjamin D Solomon 1

1. Introduction

Clinical genetics and genomics continues to evolve – for many in the field, this is a major reason why the field is so enjoyable (Slavotinek & Solomon, 2020; Solomon, 2021b). Since the turn of the millennium, milestones like the initial sequencing of the human genome, dramatic changes in sequencing technologies, and the introduction of artificial intelligence (in many different ways) have upended the field and offered fascinating new insights and directions (Biesecker & Green, 2014; Gurovich et al., 2019; Hsieh et al., 2022; Jaganathan et al., 2019; Kingsmore et al., 2019; Lander et al., 2001; Poplin et al., 2018; Zhang et al., 2022). Though difficult to predict, I think that most of us in the field feel that rapid change will continue to be inevitable, though the way in which this change manifests can be complex and difficult to forecast.

With that background in mind, I had several recent experiences that led to this project. First, I participated in an interesting exercise discussing a controversial subject (sequencing healthy newborns) (Biesecker, Green, et al., 2021). I found the exercise to be enjoyable and thought-provoking. Similarly, members of the National Human Genome Research Institute (NHGRI) released its strategic vision for the next decade (Green et al., 2020). As part of this process and document, we made ten “bold predictions” for the field of genomics and conducted a seminar series to delve further into the future of genomics. Finally, and more specifically, I attended a wonderful talk given by my colleague Dr. Les Biesecker. In general, the topic of this talk was phenotyping as applies to genetic conditions – to summarize this badly, Dr. Biesecker focused on how the approach to phenotyping has changed, and should continue to change (for some background reading, I reference some of his interesting works in this area) (Biesecker, Adam, et al., 2021; Hennekam & Biesecker, 2012; Johnston et al., 2015).

These experiences kept coming back to me – perhaps, given pandemic-related restrictions and other events in the last several years, I had a bit more time to think than usual. A few months after the lecture, I was evaluating a patient (a research participant) with a suspected genetic condition. I was with several trainees, and we discussed aspects of the physical examination. It made remember about training stories my mentors had shared with me. Some, to be honest, were genetics versions of the medical hazing rituals we can likely all recite, along the lines of, “I had to mouth pipette uphill in the snow both ways while counting fingertip whorls in the dark.” Despite the grim (and sometimes funny) stories, I compared what and how I was trying to teach the trainees to what I had heard about from my mentors, who went through training decades earlier.

Many of us in the field think about the towering contributions of David W. Smith, who described dysmorphology in a seminal 1964 paper as “the study of, or general subject of abnormal development of tissue form” (Smith, 1966). It is staggering to consider his and his colleagues’ contributions when he first defined this term, long before the underlying causes of so many genetic conditions were known, and before so many genomic (and other ‘omic)_approaches were possible. With today’s exploding technological tools and all the other changes in healthcare (e.g., the recent growth of telemedicine and remote work in general engendered by the COVID-19 pandemic) (Campbell et al., 2022; Miller et al., 2021; Shur et al., 2021; Snir et al., 2021; Tise, 2021), I wondered more about how the specific area of dysmorphology might be expected to continue to change, whether for the better (e.g., more efficient and optimal patient care and communication) or the more dystopian (e.g., removal of the human element of compassion and empathy from medicine).

With that in mind, I reached out to others in the field to solicit their feedback on the question, which might be summarized as: “What will be the practice of dysmorphology look like in ~10–20 years?” I intentionally included geneticists who I felt (subjectively!) had specific expertise and interest in this area, including individuals from different geographic regions, and who are at different career stages. Though I tried to limit the word count for each piece, I encouraged the authors to take the discussion in whichever direction they felt would be most interesting and valuable, with the understanding that not all possible topics or viewpoints would be able to be fully covered in this format. (As the areas discussed have their own vocabularies and acronyms, I have included a list of abbreviations used at the end of the combined pieces in addition to defining terms in the text as appropriate.)

I expect you will enjoy these pieces as much as I did. I feel that they offer different – sometimes overlapping and agreeing, sometimes not – lenses onto this question. These lenses naturally reflect the authors’ different experiences, interests, and clinical and research practices. I also appreciate the authors’ candor about challenges and opportunities in our field, including pointing directly to areas where change is sorely needed (Tekendo-Ngongang et al., 2020). Lastly, I want to emphasize that these lenses certainly do not reflect all possible thoughts about this question, and should not be taken to represent the “correct answers” or impart the idea that any of us has an accurate crystal ball – I feel that these pieces might best be viewed as a starting point for others in the field to ponder.

Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Margaret P Adam 1

2. The future of dysmorphology

In the early 2000s, when the first reports that the human genome had been sequenced were published (Lander et al., 2001; Venter et al., 2001), some genetics professionals worried that this might render our services obsolete. Some predicted that a genotype first approach would obviate the need for a geneticist to provide clinically relevant phenotype information to aid in diagnostics. This could not have been further from the truth, as geneticists realized that this approach requires improved phenotyping and provides opportunities for targeted, in addition to supportive, patient management strategies.

While in the past a dysmorphologist might spend time thoroughly phenotyping a patient to aid in genetic testing choices or in diagnosing syndromes for which a known genetic cause was lacking (Niikawa et al., 1988), the role has begun to shift to the requirement for deep phenotyping to aid in the interpretation of nontargeted genomic testing results. For example, the return of a result of a heterozygous pathogenic variant in KMT2D may mean a patient could have Kabuki syndrome (KS) or a completely unrelated condition characterized by choanal atresia, brachial sinus abnormalities, hearing loss with external and internal ear anomalies, interstitial lung disease, hypoplastic or absent nipples, and variable developmental delay (Baldridge et al., 2020; Cuvertino et al., 2020). The dysmorphic features are different from KS and a dysmorphologist will easily be able to distinguish between these conditions, even in neonates. Furthermore, it is not uncommon for a well characterized condition, such as Mowat-Wilson syndrome, to be diagnosed on genetic testing (for pathogenic variants in ZEB2 that lead to predicted haploinsufficiency), whereas the finding of a missense variant in ZEB2 requires expert input from a dysmorphologist to aid in the correct interpretation of such variants. In fact, many genetic testing laboratories report the finding of a pathogenic variant in GENE as leading to “GENE-related disorder” (without a phenotype designation) even if there is only one known phenotype associated with pathogenic variants in that gene. There is a reason that clinical genetic testing laboratory reports typically include a phrase such as “clinical correlation is recommended.”

But is it possible that facial recognition or other artificial intelligence diagnostic support tools could fill the role of a dysmorphologist? Most of us who have used the current facial recognition software on ourselves or our friends are all too familiar with the list of genetic syndromes that such technologies generate, despite the fact that a person may not actually have a dysmorphic genetic syndrome. Currently this type of diagnostic support tool requires the oversite of a person familiar with the proposed conditions, many of which can be easily ruled out. Furthermore, dysmorphology is more than just facial recognition. Although about 70% of dysmorphic features are found on the face and hands (Adam & Hudgins, 2003), such software typically does not include imaging of the hands. Other features require input from the user, such as the finding of a supernumerary nipple, which is often missed by the less observant provider and even the parents of an affected child. As such, it is unlikely that such software by itself will supersede the need for a good dysmorphology examination, even in the next 10–20 years. Therefore, there will still be a need for dysmorphologists to bridge the gap between pure genetic testing results and predicted clinical consequences. The main difference may be that the dysmorphology exam occurs after genetic testing, instead of prior to it.

It is also important to realize that not all dysmorphic and/or malformation conditions are due to simple genetic causes or single gene disorders that are easily identified on genomic testing. For example, diagnosing teratogenic conditions such as diabetic embryopathy, or a person with a recurrent constellation of embryonic malformation, requires the input from a dysmorphologist (Adam et al., 2020; Adam et al., 2009). While new biomarkers or other potentially novel diagnostics may become available in the future that can aid in the diagnosis of such non-Mendelian conditions, a dysmorphologist will likely be required to ensure a proper diagnosis.

Lastly, the field of dysmorphology, which has traditionally been one of primarily diagnostics, will transition to address more management issues. Targeted therapies, including gene therapy and pharmacogenomic therapies, requires a correct diagnosis. Dysmorphologists may become the main gatekeepers for ensuring appropriate targeted treatment options are implemented (Wenger et al., 2020). In summary, the role of the dysmorphologist will likely transition to include deep phenotyping after genomic testing and a greater role in patient management, including identifying targeted therapies in addition to supportive therapies.

Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Katta M Girisha 1

3. The future of dysmorphology and India

Dysmorphology is a trick that is difficult to learn and teach. I cannot easily describe how I acquired this skill and it is not easy to pass it on to my fellows. Mostly, I would admit that I learned this skill by seeing patients with syndromes, pursuing diagnostic handles in literature and recognizing them again. I am never confident of this skill though we believe I am improving with age. Even today, diagnosing a syndrome is a great challenge and a satisfying experience and will continue to be so in the near future. A slow learning curve makes an experienced dysmorphologist more valuable and sought after. Luckily, common syndromes will constitute most of our practice and rare syndromes, though there are too many for any one person to fully learn, are rare enough for us to be tolerant of embarrassing tests of our limited skills.

Artificial intelligence and machine learning will create new tools and will help both experts and beginners

We already know the impact of tools powered by artificial intelligence and machine learning in our field (Hsieh et al., 2022; Hurst, 2018). These are enabled by the standard descriptions of anomalies and dysmorphic features that can be easily used by computational tools (Allanson et al., 2009; Robinson et al., 2008). While we are comforted by the not-so-accurate “hits” in these tools, they are likely to be more precise and helpful in the days to come. These tools will make diagnoses easier and faster and will be more reliable in the hands of an experienced dysmorphologist.

Genomic tools will be more powerful but reverse phenotyping will be the norm

Next generation sequencing has improved diagnostic capabilities and enabled discovery of the causes of dysmorphologies. We have all witnessed and are overwhelmed by the pace of identification of new syndromes and their genetic bases (Bamshad et al., 2019), some subtle (Harms et al., 2017) and some striking (Gordon et al., 2017). We expect new syndrome delineation to slow down, and continue – albeit at a slower pace - for a long time, as rarer and familial syndromes continue to be discovered.

We would particularly expect trio exome sequencing and trio genome sequencing to be used for all patients as part of the evaluation of dysmorphism as the first-tier test. Algorithms for analysis of genome data are likely to improve further and replace chromosomal microarray for copy number analysis. Proteomic and transcriptomic analyses will be widely used for solving the undiagnosed syndromes (Hartley et al., 2020). Added to this will be databases of normal variants (gnomAD) (Karczewski et al., 2020), population specific datasets (Fattahi et al., 2019; “The GenomeAsia 100K Project enables genetic discoveries across Asia,” 2019; Kausthubham et al., 2021; Zhang et al., 2021) and disease-causing variants (LVOD and ClinVar) (Fokkema et al., 2011; Landrum et al., 2018), all of which will improve further and make life and variant prioritization easier for genome analysts. The ability to analyze human genomes and interpret the rare and common variants will bring implementation of “human genome on a chip” (integrated to electronic health records) a step closer. Despite these advances, the newly diagnosed syndromes will continue to need clinical validation by a dysmorphologist.

Do “normal” people have recognizable faces and syndromes?

We all share our facial features with our parents and relatives and pass them on to our offspring. We also recognize facial similarities between some individuals occasionally and immediately connect the individuals who share them (“did I meet him/her earlier?”, “I thought you are related,” or “separated at birth?” kind of expressions). It is expected that common genetic variants determine facial morphology and individuals with normal intelligence might be said to have “other syndromes” – this is different than what we currently think about when we talk about a genetic syndrome now. Recent research has shown that the genomic variations underlie facial features and brain size (Naqvi et al., 2021). Thought it may not be immediately medically relevant, curiosity might indeed lead to more mapping of such traits (such as specific morphologies of the noses, eyes or ears) in the future.

Indian scenario in the next decade

India is lagging behind most advanced centers by at least a decade but is catching up rapidly. Relative to the population size, we have disproportionately fewer clinical geneticists, dysmorphologists and genome analysts, and these rare syndromes are often neglected.

While the future dysmorphologist will be empowered with new tools – the adaptation of some of which will be grudging – our natural intelligence (the human brain) and camera (the eyes) will continue to occupy a revered corner of genomic medicine.

Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Judith G Hall 1,2

4. Dysmorphology moves on to natural history

These articles are meant to describe how phenotyping will be practiced in 10 to 20 years. As an individual who tried to help define terms and standards over many years, I am troubled by the fact that medical genetics fellowship training seems to put little emphasis on correct physical diagnosis and defining the natural history of various disorders. It appears that ordering molecular diagnostic tests and following algorithms rather than learning to think through disease processes will become much of the practice of dysmorphology in the future.

The efforts in Medical Genetics to define morphologic terms and measuring various physical features in the 90s led to improved physical diagnosis prior to the advent of genomics (Biesecker et al., 2009; Gripp et al., 2013; Hall et al., 2007; Hennekam et al., 2013). However, defining specific disorders has moved from a complex of physical features to a specific variant in a specific gene. The availability of next generation sequencing both of exomes and whole genome sequencing has inverted the process of the diagnosis of patient (Grody, 2019). Indeed, when one has a rare disorder and identifies the gene variant, one searches all over the world to find someone else with the same disorder (Bastarache et al., 2018). Then, the physical and behavioral description can emerge (Shalev & Hall, 2004), the timing of development of various symptoms can be defined, and gender specific differences can be described. Even in environmentally caused disorders, such as those related to infections or drugs, the genome often plays a part in susceptibility. Indeed, ethnic differences are just beginning to be defined.

A newer, but important to families, part of the phenotype that is now being described is the natural history of the disorder; what happens in each organ system over time (Hall, 1985); including the behavioral phenotypes. Unusual behaviors, the types of developmental disability, and neurologic complications are being better described. In rare diseases and disorders, the parent support groups are playing a major part in wanting the natural histories to be defined, so that they can plan for the future. Searching electronic records, facial recognition, and computation of physical features require special training and expertise, but hopefully will play a role in defining and diagnosis in the future (Hallowell et al., 2019; Solomon, 2021a).

Indeed, the inclusion of families in planning research has led to new and interesting issues. “Nothing about us without us” leads to a new set of concerns and a change in vocabulary: “funny looking kid”, “handicapped”, and “mental retardation” are terms which are not used any longer.

In order for the natural histories of specific variants to be defined, a great deal of accurate historical data needs to be accumulated and it takes a surprisingly long time and large amount of effort to accumulate. Registries of specific disorders are often available for children but to have long term adult information is what is really needed for the proper description of natural history. That is the kind of information families wish to have! It will identify at what age one needs to do screening and consider possible interventions.

Unfortunately, neither granting agencies nor insurance companies want to pay for the additional time required to define, describe, and record natural history. The granting agencies seem to think that it is part of a clinical evaluation, but those who pay for clinical evaluation limit the amount of time that they will reimburse.

In addition, what is really needed is the time in development, sex specific, tissue specific gene expression and thereby the age and sex specific of pathways that are involved in the function of that tissue. In general, adults are studied for the proteins available in various tissues, and little work has been done to understand the shifting of physiology over time in various tissues. The changes, which are part of the process of normal human growth and development, have undoubtedly evolved over eons as part of selection of the fittest. Each change had made human survival more likely, for instance, from embryonic, to fetal, to adult hemoglobin expression. That defining of natural history then supplies information that can be used as therapy in rare disorders.

This field of Dysmorphology (and Clinical/Medical Genetics) will continue to evolve, as we have seen in these last few years (Biesecker, Adam, et al., 2021). It reminds me of efforts in earlier stage of Dysmorphology to have standard terminology for elements of morphology and developed standard weight and height curves for rare disorders and more recently to have them available for various ethnic groups. This is a field that will continue to evolve as the epigenetic control of pathways are uncovered and age specific, sex specific tissues are defined, even if the genetics fellows are not thoroughly trained in physical diagnosis.

Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Anna CE Hurst 1

5. The evolving role of dysmorphology

This is an incredible time to work in rare diseases. We have access to exome and genome sequencing (ES/GS), functional studies, improved databases, artificial intelligence, facial recognition software. But the most essential tool we have is the dysmorphology physical exam.

New tools don’t replace the role of a dysmorphologist – they bolster it and highlight new areas where a geneticist can help make sense of it all. Dysmorphology is a tool, and while it is no longer the sole resource we have in our diagnostic arsenal, we should not diminish its relevance in light of newer advances. Craniofacial and physical subtleties underly and explain mechanisms of malformation and give insights into the “how and why” of genetic conditions.

The rapid integration of ES/GS technologies is transforming medical genetics, but dysmorphology still remains the foundation of making a diagnosis and characterizing new syndromes. While the timing of the greatest use of dysmorphology is shifting from pre-test selection toward post-test interpretation, early integration of dysmorphology is still relevant, as it can led to cost-effective test selection (Hurst & Robin, 2020). In other stages of diagnosis, dysmorphology is needed for test interpretation for formal diagnosis and description of conditions.

New technologies like ES/GS are additional tools to gather evidence in the quest to make a diagnosis, but ES/GS isn’t always the best test for every patient. It remains more expensive than focused panels or microarrays, and often a patient’s physical exam and history can direct a geneticist to a narrow differential diagnosis. From personal experience working with research GS performed in parallel with clinical testing, there are cases where a focused clinical testing approach leads to a molecular diagnosis faster than GS. And there are cases where physical exam leads to a clinical diagnosis and directs management before testing is even started.

As the field seeks to expand genetic testing throughout medical care, we must be conscious of the financial and time costs of our testing efforts. Additionally, some diagnoses require a unique testing strategy, such as methylation or fibroblast-based studies, and a dysmorphologist can often recognize when there is a different “best test” or when testing is likely to be non-diagnostic.

When faced with non-diagnostic or negative testing, a dysmorphologist can recognize when features establish a clinical diagnosis. Or, when faced with a positive test, a geneticist may recognize that an existing diagnosis does not fully encompass all of a patient’s phenotype and a second diagnosis may be present.

When ES/GS is performed, variants are usually filtered using phenotypic terms. If a lab relies on HPO terms in the variant filtration pipeline, then it is up to the referring clinician to provide the most accurate description of the patient possible. Without a clear phenotype, a genotype is insufficient.

Despite our best efforts at variant classification, many ES/GS findings remain variants of uncertain significance (VUS). This is the time when the dysmorphologist’s exam is most valuable. We must consider the role of a gene in craniofacial development, features reported in the literature, and an unbiased assessment of the patient. We must also assess the patient in the context of relatives’ genotypes and phenotypes, while considering if their features could be due to shared familial or ancestral backgrounds. Dysmorphology bridges patients and the laboratory, making sense of phenotype together with a genotype.

For some neurodevelopmental disability syndromes, the distinctive features are craniofacial dysmorphisms, so VUSes in those genes are more likely to be benign if such features are absent. Of course, this is complicated by variable expressivity or reduced penetrance, and strong dysmorphology skills can also help in assessing those aspects of genetic conditions.

A favorite question among geneticists is, “What is a syndrome?” Are syndromes classified by a shared molecular origin or a consistent phenotypic description? For novel syndromes, determining if a candidate gene leads to shared features in a patient cohort requires a careful assessment. Dysmorphologists are poised to distinguish between allelic disorders, such as when there are differences between gain-of-function and loss-of-function variants in the same gene.

Additionally, dysmorphic findings in patient cohorts may be due to a central mechanism of action of the gene in question, and understanding mechanisms underlies the development of future treatments for many conditions.

As a field we must recognize the value in all the tools at our disposal, not forgetting that a detailed history combined with a keen eye and focused physical exam remains central to rare disease diagnosis.

Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Shahida Moosa 1,2, Peter M Krawitz 3

6. The future of dysmorphology: merging human and AI abilities

When we started our training in medical genetics, Germany had already begun adopting genomics, while South Africa lagged behind. The emphasis on the clinical training in medical genetics in our respective countries reflected this. South African geneticists were trained to be classical or “forward” dysmorphologists, as the lack of genomic testing meant that a clearer clinical differential diagnosis would be relied heavily upon to guide management and care of patients. In contrast, German geneticists were trained to be both “forward” and “reverse” dysmorphologists, with the emphasis on reverse phenotyping increasing with time as more genetic testing became available. The prediction back then was that dysmorphology would become an obsolete skill due to increased genomic testing.

As medical geneticists who are still active dysmorphologists, now training the next generation, we find ourselves questioning the initial prediction and wondering about the future of classical dysmorphology. Where should the emphasis now lie? Genomic testing in Africa is still in its infancy, but should that necessarily mean that we maintain the status quo? Similarly, is it enough for a German trainee to recognize that the patient has dysmorphic features, and then wait for the panel/exome/genome result and perform reverse phenotyping?

We argue that the above should be challenged. Classical (forward) dysmorphology still has a role to play but that perhaps the expectation of the dysmorphologist to arrive at a diagnosis should be tempered. It is an imperfect science, even for the most experienced practitioners. With the growing number of syndromes, it is impossible for even expert dysmorphologists to keep abreast with the literature. Often, new syndromes are delineated in a handful of patients, making it difficult for anyone to become an expert on all syndromes. Furthermore, despite the falling costs of sequencing, the amount of data produced is still very large and we feel that all means should be utilized to narrow the search space for the underlying causative variant. This includes, but is no longer limited to, excellent “forward” dysmorphology, including medical photography and the use of human phenotype ontology (HPO) or clinical descriptive terms. This requires the human eye to recognize the features, but we increasingly recognize that the human alone is not enough.

We believe that artificial intelligence (AI), and so-called next generation phenotyping (NGP) tools, which combine the human and the machine “dysmorphologists”, is the way of the future no matter the geographical location. Our ideas on what “dysmorphology” will look like in the future is based on tools already available, those in development and how the field is moving to achieve a beneficial balance between human and machine. This collaboration is essential in solving age-old challenges, like the lumping vs splitting debate, as well as in the accurate delineation of new syndromes and clinical spectra of disorders.

Proof of concept and success in the clinical setting has already been shown by tools like PEDIA (Hsieh et al., 2019) and others, which combine computer-assisted analyses of facial images and clinical features can increase diagnostic yield. Newer NGP tools like GestaltMatcher (Hsieh et al., 2022) have shown that the AI-driven tools are very successful at matching patients, especially those with ultra-rare disorders and even novel disorders. These examples show that the incorporation of NGP tools far exceeds the expert human dysmorphologists’ abilities in isolation. But just as the human’s ability is dependent on their experience, so the AI is dependent on what data it has been trained on. Both are imperfect in isolation. Merging their abilities maximizes benefit.

In future, tools which incorporate other images will become popular; the list of possibilities is endless, including anything from skin lesions, to plain film radiographs, to retinal images or electroencephalograms. We envision that these tools will also be used in other applications: monitoring response to therapies, which may influence or change the dysmorphic features, or using neonatal facial images and artificially aging them to reveal the dysmorphology, without needing to wait for the child to grow.

However, most NGP tools still do not have enough representation from many ethnicities across the globe, making them less accurate when used on African patients, for example. With the newer tools in development, this will change. Which brings us to the point of accessibility and broadening the user base of the NGP. These tools can and should be made accessible to any healthcare worker with a smartphone, who does not necessarily have extensive dysmorphology training. We know that the majority of patients cannot access genetic services, especially in resource-restricted environments. Making these tools intuitive and easy to use, potentially and importantly, would broaden access to the broader medical community caring for the patients who would benefit from a dysmorphology assessment. There is a huge need in Africa, for example, that these tools move out of the realm of medical genetics (of which there are still far too few centers on the continent), to increase accessibility. We foresee pediatricians, family physicians and community nurses even in countries without genetic services, using these tools to help their patients.

We will continue training expert “forward” and “reverse” dysmorphologists, who recognize that they cannot work optimally in isolation. A happy collaboration between human and AI is needed to merge and optimize their combined abilities.

Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Shubha R Phadke 1

7. Changing the face of dysmorphology: different beauty

Beauty lies in the eyes of the beholder. As physicians who regularly take care of people and families affected by genetic conditions, we have to consider the terms we use, and the way we think about people with these conditions: is it appropriate to label different facial features of patients with genetic disorders as “dys-morphic”? This label, and the thought process behind it, seems to indicate a negative value judgment. Children with genetic conditions may have features that are medically important to the eyes of clinical geneticists, as they help us make clinical diagnoses. But we must be careful about “dys”-based terminology. It is time we remove “dysmorphic facies” from our glossaries and simply refer to the overall facial phenotype. Or, we can think of some other term, as we use words like “differently abled.” For differently abled people, the Indian prime minister has introduced the word “divyang.” This word is a combination of two Sanskrit words: “divya,” which means divine and “ang,” which refers to an organ or a body part. We can think of such a word for unique characteristic facial phenotypes of genetic syndromes as “swaroop”: “swa” (self), and “roop” (the way a person looks), their facial features or self-identity. This emphasizes the uniqueness of a person’s facial phenotype without labelling it as abnormal. The word “swaroop” also means a lover of beauty and indicates positivity associated with the term. Overall, it is time to consider swapping dysmorphology for “swaroopology” to help ensure dignity and respect for all people.

Another issue about facial phenotypes, in the new era of artificial intelligence (AI), is the changing role of clinical geneticists. When I started my career in medical genetics in the 1990s, syndromology was my area of interest as it did not require sophisticated laboratory support, such as needed for inborn errors of metabolism. Everyone learns about the syndrome diagnosis but some are better at it than others. As facial recognition software is showing great improvement, it seems sometimes that the clinical expertise in swaroopology may become redundant. This is unlikely, though. Reverse phenotyping will always be necessary and will become more and more important. The clinical geneticists of the new era need to be astute and highly observant clinicians as the genomic tools are able to make molecular diagnoses of syndromes where facial phenotypes, though characteristic, have very subtle phenotypes. And the need to be able to make clinical diagnoses for many common malformation syndromes based on the pattern and facial phenotypes should not be underestimated. With genomic techniques, we are now diagnosing many rare syndromes. Re-examining and comparing the facial features of molecularly diagnosed individuals makes us able to better perceive and register their characteristic feature (in our own brains, not just in AI tools) in retrospect. Examples include the shape of the nose of people with Koolen de Vries syndrome, the lower lip in people with Kleefstra syndrome, the upper lip of people with Smith Magenis syndrome and straight eyebrows in people with chromosome 1p deletion syndrome, to name a few (Gupta et al., 2016; Kandasamy et al., 2014). With experience, the clinical diagnosis of syndromes with subtle facial phenotypes may become increasingly possible. The new technologies should thus sharpen the clinical skills of the pre-AI generation of clinical geneticists and be complementary. I wonder if the next generation will miss the pleasure of having a clinical diagnosis confirmed by a genetic test!

As genomic data may be available for all newborns, the first clinical evaluation involve reverse phenotyping, with the bias of a probable genetic diagnosis. The admixture of populations world over is sure to pose problems of phenotyping and genotyping. Diagnosis based on physical examination will become easier and the same AI tools used for diagnosis will be useful in documenting the clinical phenotypes that are needed by clinicians for further understanding and synthesizing data (Hall, 2003). The clinicians’ eyes and mind will be able to leverage the power of image recognition technologies. Virtual platforms like telemedicine will also become very important in dysmorphology diagnosis. The availability of expert dysmorphologists who have seen a number of cases of any rare syndrome will be more easily available for second opinions and advice. Advances in technologies will in return train our minds and we will learn to consider the opinions of software tools, just as we benefit from our colleagues in difficult cases. With electronic health records, more information about natural course of various syndromes, including images and videos at different ages, should become available in the next decades if we start making systematic efforts. These will be helpful for developing and testing therapies and for families of patients who seek to know more about the condition and the prognosis.

As we are trying to encompass all people and give equal opportunities to all, the time is right for everyone, who should all be proud of their “swaroop.”

Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Cedrik Tekendo-Ngongang 1

8. The future of dysmorphology: a global perspective

While dysmorphology is part of clinical education and medical genetics practice around the world, its contribution to patient care is likely not similarly regarded across healthcare systems or countries. Recent observations suggest an inverse relationship between access to clinical genomic testing and judgements about the worthiness of dysmorphology practice in various healthcare systems; the less genetic and genomic tests are available and accessible to the public, the more clinicians would rely on their dysmorphology skills in routine clinical practice (Tekendo-Ngongang et al., 2014). The enduring imbalance in access to medical genetics services between high- and low-income countries requires advocacy for a less globalized approach in efforts to hypothesize the future of dysmorphology (Abacan et al., 2019). Accordingly, for the coming decade or two, it will not be unrealistic to forecast distinct trends in the practice of dysmorphology, which will closely align with the level of penetration of genomic technologies in healthcare systems across the globe. In addition, considering effective applications of artificial intelligence (AI) methods to dysmorphology lately, one may expect alterations in the practice of dysmorphology in the short term.

Anticipated trend in more advanced healthcare systems: interest in dysmorphology practice will likely fade away, but will not completely disappear

Clinical genomic tests are accessible to a substantial fraction of healthcare providers and the public in high-income countries. These tests are predicted to experience an exponential qualitative and quantitative expansion in the next decade (Green et al., 2020). Cheaper and readily available genome wide testing such as chromosomal microarray, exome or genome sequencing (which may require less detailed initial phenotypic information) may legitimately reinforce disagreements on the need for fine dysmorphology skills both in education (e.g., clinical genetics and genomics residency training) and routine clinical practice; such a scenario could lead to gradual decline in interest in traditional dysmorphology practice and education. However, considering that detailed and accurate phenotypic information will more than ever be required to facilitate interpretation of clinical genomic testing (especially in cases of “reverse phenotyping”), there will still be room for application of dysmorphology skills in routine medical practice.

It is worth mentioning that lessons from the history of medicine suggest that there can be a downside to technological advances in general and the advent of modern diagnostic modalities in particular; while data from dysmorphology assessment will be needed to help interpret genomic testing, one could rationally expect that rapid advances in genomic technology may lead to loss of skills in terms of precisely describing a clinical phenotype, just as the loss of the detailed descriptions of clinical anatomy occurred with the advent and advances of diagnostic radiologic modalities.

If the next generation of clinical geneticists must be competent not just at ordering genomic testing and interpreting genomic variants, being able to accurately describe a clinical phenotype and formulate a sound differential diagnosis will remain essential. A predictable challenge will be ensuring that the training of future geneticists in dysmorphology is maintained across genetics and genomics training programs while still adjusting for ongoing advances in genomic medicine.

Anticipated trend in less resourced healthcare systems: interest in dysmorphology practice will likely remain high

Access to medical genetics services and genomic technology remain a real issue in many low-income countries – two common characteristics of these areas are dearth of resources, and underrepresentation of their populations in medical genetics and genomics materials and databases (Green et al., 2020). While dysmorphology is, despite the scarcity of trained dysmorphologists, routinely practiced to some extent in several low-income countries, the routine availability of basic genetic and genomic tests is yet to become a reality in their health systems. Nevertheless, efforts to bridge the genomic technological gap between these and the more advanced healthcare systems (see: https://h3africa.org; https://latinamericangenomicsconsortium.org) (Wonkam, 2021), as well as the rising number of genetic and genomic research being conducted (Cavalcanti et al., 2020; Ekure et al., 2021; Mistri et al., 2019) suggest that integration of clinical genomic testing into routine clinical practice can be projected in these parts of the globe in the coming two decades or so – but unlikely before that time. Meanwhile, healthcare systems will continue to rely on practitioners’ clinical skills, including fine dysmorphology skills, to mitigate the deficit in molecular testing.

The growing impact of artificial intelligence (AI) in medicine: digital dysmorphology will gradually supplement the traditional practice of clinical genetics, both in advanced and less advances healthcare systems

The last decade has witnessed an unprecedented increase in interest in applications of AI to medicine and biomedical sciences, and a surge in AI translational research initiatives in medical genetics and genomics (Duong et al., 2022; Poplin et al., 2018). Specifically, machine learning methods are being applied successfully for diagnostic processes involving pattern recognition, and the increasing use of computer-assisted facial images processing for genetic syndromes recognition is a perfect illustration of how machine learning methods are being applied to dysmorphology (Hsieh et al., 2022; Kruszka et al., 2019). The ability to utilize machine learning algorithms to accurately classify or label patients with specific genetic syndromes based on their subtle phenotypic patterns (a process also known as “digital dysmorphology”) could well be regarded as a revolution in dysmorphology. Rationally, one may anticipate such a compelling diagnostic approach to be gradually adopted by genetic and non-genetic clinicians as an adjunct to conventional dysmorphology, especially in settings with limited genetic workforce and clinical genetics expertise. That said, the universal shortage of physician geneticists (Jenkins et al., 2021) suggests that expansion of digital dysmorphology will remodel the practice of clinical genetics in general and dysmorphology in particular to varying degrees in both high- and low-income countries.

Am J Med Genet A. 2022 Dec 9;191(3):659–671.
Tara L Wenger, Chin-To Fong 1,2

9. The future of dysmorphology: advocacy, computational enhancement, and genomics

Dysmorphology has traditionally leveraged the innate ability to detect minor differences in faces to predict genotypic differences, which in turn predict the broader phenotype. Knowing the natural history of a specific disorder allows us to predict and potentially disrupt the typical medical consequences with appropriate interventions. This approach is very helpful for conditions in which a specific genetic change results in a strong phenotype that includes a recognizable facial phenotype.

There are several limitations to dysmorphology including: 1) genetic disorders without a clear facial phenotype; 2) limitations in understanding of what dysmorphic features are apparent at different life stages from prematurity through adulthood; 3) textbook bias for manifestations of genetic syndromes in predominantly white infants and older individuals; 4) implicit bias engendered by limited observer experience for facial features of diverse population groups, leading to decreased dysmorphology detection rate among non-white patients; 5) disparate access to a dysmorphologist.

Increasing access to exome and genome sequencing has allowed for genetic diagnosis in a subset of individuals that would not have received a diagnosis based on dysmorphology alone. However, genomic testing has its own unique limitations, including but not limited to: 1) sampling of appropriate tissue for mosaic disorders; 2) limited ability to measure epigenetic changes; 3) features that occur as the result of exposure to a teratogen; 4) genetic conditions that are not evaluated well by this technology (e.g. methylation defects, trinucleotide repeat disorders); 5) difficulty interpreting variants of uncertain significance.

While identification of an increasing number of genetic disorders can be made with genomic testing alone, many infants with birth defects or dysmorphic features remain without a clear genetic explanation. For affected infants with negative genomic testing a dysmorphologist can use facial and morphological differences to predict broader phenotype and diagnostic categories (e.g. recurrent constellations of embryonic malformation such as VACTERL, craniofacial microsomia).

Genomic testing as dysmorphologist-extender

Dysmorphologists leverage experiential training of the highly evolved human capacity for recognizing, memorizing and categorizing facial details and patterns with studied understanding of genomic testing to make diagnosis of rare genetic disorders. However, there is a massive shortage of geneticists, with most wait lists that are many months long, and most NICUs and birth hospitals not having access to a dysmorphologist, creating a service gap that is every widening as the number of diagnosable genetic disorders continues to rise. Therefore, in the future, we predict that genomic testing will be ordered by a broader range of physicians, with assistance from geneticists and genetic counselors to help interpret the findings. This approach will reduce the overwhelming pool of individuals who have been referred to a smaller number that were not diagnosable by broad testing, and therefore require the skillset of the dysmorphologist.

Computational science as dysmorphologist-enhancer

Looking forward, how can we best leverage genomics and computational science to better train dysmorphologists and enhance identification of genetic disorders? First, artificial intelligence. Facial appearance can be reasonably constructed using DNA, which has largely been used in forensic cases. A major limitation in dysmorphology is the ability of geneticists to recognize dysmorphic features in individuals from ethnic background they have limited experience with. In contrast to existing technology that averages facial features to reproduce a gestalt, we predict that in the future technology will use an individual’s specific as well as background DNA sequence data that will allow one to superimpose the phenotypic effect of rare deleterious variants against common population-specific variants, allowing an improved sensitivity of diagnosis in patients from minority populations. As this technology develops, it could allow for a dysmorphologist to better appreciate what is difference about the appearance of that individual compared to what was predicted based on their genetic background. Similarly, age progression technology could be used to predict what an individual’s face could be predicted to look like at different ages. Any discrepancy between the expected and observed facial phenotype may reflect the impact of the deleterious variant on the aging process itself.

Dysmorphologists as advocates

In the future, we predict that dysmorphologists will be critical in advocating for better testing policies, improved coverage for preimplantation genetic testing for familial genetic disorders and improved protections for life and disability insurance for individuals with genetic disease. At a time when the public at large has lost faith in scientists, genetics has seemed to retain trust. Widespread awareness of genetic disease and direct to consumer genetic testing has improved in recent years. Moreover, dysmorphologists will be needed to advocate for the right of physicians to counsel women about their reproductive choices in the setting of fetal anomalies.

Dysmorphologists are also needed to educate payors and lawmakers regarding the necessity of coverage for genetic counseling services. This will be even more necessary as there is an inevitable disruption of the genetic testing workflow with wider adoption of early broad genomic sequencing, when patients may need to meet with a genetic counselor to discuss results from genetic testing prior to their evaluation by a dysmorphologist. The conventional workflow has permitted lack of coverage policies for genetic counseling services in some situations because of the convention of having a physician geneticist attached to each visit.

Dysmorphologists are uniquely poised to advocate for the rights of physicians to counsel women. While clinical geneticists are able to test and counsel about conditions that can be detected on genetic testing, dysmorphologists are best suited to prognosticate in the absence of molecular confirmation. Many multiple malformation syndromes including Recurrent Constellations of Embryonic Malformation rely on a clinical diagnosis. The art of prognostication in the absence of a molecular diagnosis is needed during pregnancies identified to have an affected fetus, before molecular testing or when prenatal testing is not possible. There are no shortcuts around a dysmorphology assessment in the absence of molecular testing. Dysmorphologists are also the best positioned to identify when a molecular diagnosis does not account for all features, or when an abnormal molecular result does not match the phenotype of the patient. This is perhaps the best illustration that complex assessment, decision-making and prognostication should be left between patient and their physicians, and that individuals without a deep understanding of developmental biology and limitations of testing should not be involved in restricting the rights of others.

RESOURCES