Crustal anatexis in collisional orogens has great bearing on geochemical differentiation of the continental crust. However, it is often uncertain what kind of crustal rocks were partially melted for felsic magmatism at convergent plate boundaries. To address this issue, a combined study of in-situ monazite U–Th–Pb ages, in-situ monazite and allanite SmNd isotopes and whole-rock SmNd isotopes was carried out for Higher Himalayan metamorphic rocks and leucogranites in the Himalayan orogen. Metapelite, metagreywacke and granitic gneiss are the dominant constituents of the Higher Himalayan Crystallines, and they experienced upper amphibolite to granulite facies metamorphism at ca. 26–13 Ma. Although the three rock types show consistently negative εNd(t) values at t = 20 Ma, their Nd isotope compositions become less and less enriched from metapelite through metagreywacke to granitic gneiss. The metapelite has the lowest εNd(t) values of −19.9 to −15.7, the metagreywacke has intermediate εNd(t) values of −17.4 to −12.7, and the granitic gneiss has the highest εNd(t) values of −14.1 to −7.7. Leucogranitic magmatism occurred at ages from ca. 26 to 7 Ma, coeval with anatectic metamorphism of the three rock types during the Oligocene–Miocene. Relict zircon UPb age distributions, whole-rock trace element patterns, initial Nd isotope compositions and two-stage Nd model ages for the Higher Himalayan leucogranites are comparable with those for the Higher Himalayan metamorphic rocks, confirming that the metapelite, metagreywacke and granitic gneiss would have their compositional counterparts at structurally deeper positions to serve as the crustal sources of the leucogranites. This is also verified by forward phase equilibrium modelling for partial melting of the metamorphic rocks with respect to the differences in their composition and fertility. In addition, the leucogranites show a decreasing trend in εNd(t) values with their ages, indicating that the dominant crustal sources would gradually change from the least fertile granitic gneiss through the intermediately fertile metagreywacke to the most fertile metapelite during the protracted crustal anatexis from the late Oligocene to the Miocene. Therefore, the deep crust in the Himalayan orogen would consist of the metamorphic rocks with similar compositions to the shallow crust. The combined geochronological and geochemical study of accessory minerals and host rocks can provide the genetic link between the crustal sources and their melting products in collisional orogens. Nevertheless, the Nd isotope variation in these leucogranites may also be related to incongruent melting of the crustal sources, which has a potential to result in the Nd isotope disequilibrium between melt and residue.